Adding upstream version 2:8.2.2434.upstream/2%8.2.2434upstream

Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>

1 files changed, 2856 insertions, 0 deletions

diff --git a/src/findfile.c b/src/findfile.c
new file mode 100644
index 0000000..4f69c47
--- /dev/null
+++ b/src/findfile.c
@@ -0,0 +1,2856 @@
+/* vi:set ts=8 sts=4 sw=4 noet:
+ *
+ * VIM - Vi IMproved	by Bram Moolenaar
+ *
+ * Do ":help uganda"  in Vim to read copying and usage conditions.
+ * Do ":help credits" in Vim to see a list of people who contributed.
+ * See README.txt for an overview of the Vim source code.
+ */
+
+/*
+ * findfile.c: Search for files in directories listed in 'path'
+ */
+
+#include "vim.h"
+
+/*
+ * File searching functions for 'path', 'tags' and 'cdpath' options.
+ * External visible functions:
+ * vim_findfile_init()		creates/initialises the search context
+ * vim_findfile_free_visited()	free list of visited files/dirs of search
+ *				context
+ * vim_findfile()		find a file in the search context
+ * vim_findfile_cleanup()	cleanup/free search context created by
+ *				vim_findfile_init()
+ *
+ * All static functions and variables start with 'ff_'
+ *
+ * In general it works like this:
+ * First you create yourself a search context by calling vim_findfile_init().
+ * It is possible to give a search context from a previous call to
+ * vim_findfile_init(), so it can be reused. After this you call vim_findfile()
+ * until you are satisfied with the result or it returns NULL. On every call it
+ * returns the next file which matches the conditions given to
+ * vim_findfile_init(). If it doesn't find a next file it returns NULL.
+ *
+ * It is possible to call vim_findfile_init() again to reinitialise your search
+ * with some new parameters. Don't forget to pass your old search context to
+ * it, so it can reuse it and especially reuse the list of already visited
+ * directories. If you want to delete the list of already visited directories
+ * simply call vim_findfile_free_visited().
+ *
+ * When you are done call vim_findfile_cleanup() to free the search context.
+ *
+ * The function vim_findfile_init() has a long comment, which describes the
+ * needed parameters.
+ *
+ *
+ *
+ * ATTENTION:
+ * ==========
+ *	Also we use an allocated search context here, this functions are NOT
+ *	thread-safe!!!!!
+ *
+ *	To minimize parameter passing (or because I'm to lazy), only the
+ *	external visible functions get a search context as a parameter. This is
+ *	then assigned to a static global, which is used throughout the local
+ *	functions.
+ */
+
+/*
+ * type for the directory search stack
+ */
+typedef struct ff_stack
+{
+    struct ff_stack	*ffs_prev;
+
+    // the fix part (no wildcards) and the part containing the wildcards
+    // of the search path
+    char_u		*ffs_fix_path;
+#ifdef FEAT_PATH_EXTRA
+    char_u		*ffs_wc_path;
+#endif
+
+    // files/dirs found in the above directory, matched by the first wildcard
+    // of wc_part
+    char_u		**ffs_filearray;
+    int			ffs_filearray_size;
+    char_u		ffs_filearray_cur;   // needed for partly handled dirs
+
+    // to store status of partly handled directories
+    // 0: we work on this directory for the first time
+    // 1: this directory was partly searched in an earlier step
+    int			ffs_stage;
+
+    // How deep are we in the directory tree?
+    // Counts backward from value of level parameter to vim_findfile_init
+    int			ffs_level;
+
+    // Did we already expand '**' to an empty string?
+    int			ffs_star_star_empty;
+} ff_stack_T;
+
+/*
+ * type for already visited directories or files.
+ */
+typedef struct ff_visited
+{
+    struct ff_visited	*ffv_next;
+
+#ifdef FEAT_PATH_EXTRA
+    // Visited directories are different if the wildcard string are
+    // different. So we have to save it.
+    char_u		*ffv_wc_path;
+#endif
+    // for unix use inode etc for comparison (needed because of links), else
+    // use filename.
+#ifdef UNIX
+    int			ffv_dev_valid;	// ffv_dev and ffv_ino were set
+    dev_t		ffv_dev;	// device number
+    ino_t		ffv_ino;	// inode number
+#endif
+    // The memory for this struct is allocated according to the length of
+    // ffv_fname.
+    char_u		ffv_fname[1];	// actually longer
+} ff_visited_T;
+
+/*
+ * We might have to manage several visited lists during a search.
+ * This is especially needed for the tags option. If tags is set to:
+ *      "./++/tags,./++/TAGS,++/tags"  (replace + with *)
+ * So we have to do 3 searches:
+ *   1) search from the current files directory downward for the file "tags"
+ *   2) search from the current files directory downward for the file "TAGS"
+ *   3) search from Vims current directory downwards for the file "tags"
+ * As you can see, the first and the third search are for the same file, so for
+ * the third search we can use the visited list of the first search. For the
+ * second search we must start from a empty visited list.
+ * The struct ff_visited_list_hdr is used to manage a linked list of already
+ * visited lists.
+ */
+typedef struct ff_visited_list_hdr
+{
+    struct ff_visited_list_hdr	*ffvl_next;
+
+    // the filename the attached visited list is for
+    char_u			*ffvl_filename;
+
+    ff_visited_T		*ffvl_visited_list;
+
+} ff_visited_list_hdr_T;
+
+
+/*
+ * '**' can be expanded to several directory levels.
+ * Set the default maximum depth.
+ */
+#define FF_MAX_STAR_STAR_EXPAND ((char_u)30)
+
+/*
+ * The search context:
+ *   ffsc_stack_ptr:	the stack for the dirs to search
+ *   ffsc_visited_list: the currently active visited list
+ *   ffsc_dir_visited_list: the currently active visited list for search dirs
+ *   ffsc_visited_lists_list: the list of all visited lists
+ *   ffsc_dir_visited_lists_list: the list of all visited lists for search dirs
+ *   ffsc_file_to_search:     the file to search for
+ *   ffsc_start_dir:	the starting directory, if search path was relative
+ *   ffsc_fix_path:	the fix part of the given path (without wildcards)
+ *			Needed for upward search.
+ *   ffsc_wc_path:	the part of the given path containing wildcards
+ *   ffsc_level:	how many levels of dirs to search downwards
+ *   ffsc_stopdirs_v:	array of stop directories for upward search
+ *   ffsc_find_what:	FINDFILE_BOTH, FINDFILE_DIR or FINDFILE_FILE
+ *   ffsc_tagfile:	searching for tags file, don't use 'suffixesadd'
+ */
+typedef struct ff_search_ctx_T
+{
+     ff_stack_T			*ffsc_stack_ptr;
+     ff_visited_list_hdr_T	*ffsc_visited_list;
+     ff_visited_list_hdr_T	*ffsc_dir_visited_list;
+     ff_visited_list_hdr_T	*ffsc_visited_lists_list;
+     ff_visited_list_hdr_T	*ffsc_dir_visited_lists_list;
+     char_u			*ffsc_file_to_search;
+     char_u			*ffsc_start_dir;
+     char_u			*ffsc_fix_path;
+#ifdef FEAT_PATH_EXTRA
+     char_u			*ffsc_wc_path;
+     int			ffsc_level;
+     char_u			**ffsc_stopdirs_v;
+#endif
+     int			ffsc_find_what;
+     int			ffsc_tagfile;
+} ff_search_ctx_T;
+
+// locally needed functions
+#ifdef FEAT_PATH_EXTRA
+static int ff_check_visited(ff_visited_T **, char_u *, char_u *);
+#else
+static int ff_check_visited(ff_visited_T **, char_u *);
+#endif
+static void vim_findfile_free_visited(void *search_ctx_arg);
+static void vim_findfile_free_visited_list(ff_visited_list_hdr_T **list_headp);
+static void ff_free_visited_list(ff_visited_T *vl);
+static ff_visited_list_hdr_T* ff_get_visited_list(char_u *, ff_visited_list_hdr_T **list_headp);
+
+static void ff_push(ff_search_ctx_T *search_ctx, ff_stack_T *stack_ptr);
+static ff_stack_T *ff_pop(ff_search_ctx_T *search_ctx);
+static void ff_clear(ff_search_ctx_T *search_ctx);
+static void ff_free_stack_element(ff_stack_T *stack_ptr);
+#ifdef FEAT_PATH_EXTRA
+static ff_stack_T *ff_create_stack_element(char_u *, char_u *, int, int);
+#else
+static ff_stack_T *ff_create_stack_element(char_u *, int, int);
+#endif
+#ifdef FEAT_PATH_EXTRA
+static int ff_path_in_stoplist(char_u *, int, char_u **);
+#endif
+
+static char_u e_pathtoolong[] = N_("E854: path too long for completion");
+
+static char_u	*ff_expand_buffer = NULL; // used for expanding filenames
+
+#if 0
+/*
+ * if someone likes findfirst/findnext, here are the functions
+ * NOT TESTED!!
+ */
+
+static void *ff_fn_search_context = NULL;
+
+    char_u *
+vim_findfirst(char_u *path, char_u *filename, int level)
+{
+    ff_fn_search_context =
+	vim_findfile_init(path, filename, NULL, level, TRUE, FALSE,
+		ff_fn_search_context, rel_fname);
+    if (NULL == ff_fn_search_context)
+	return NULL;
+    else
+	return vim_findnext()
+}
+
+    char_u *
+vim_findnext(void)
+{
+    char_u *ret = vim_findfile(ff_fn_search_context);
+
+    if (NULL == ret)
+    {
+	vim_findfile_cleanup(ff_fn_search_context);
+	ff_fn_search_context = NULL;
+    }
+    return ret;
+}
+#endif
+
+/*
+ * Initialization routine for vim_findfile().
+ *
+ * Returns the newly allocated search context or NULL if an error occurred.
+ *
+ * Don't forget to clean up by calling vim_findfile_cleanup() if you are done
+ * with the search context.
+ *
+ * Find the file 'filename' in the directory 'path'.
+ * The parameter 'path' may contain wildcards. If so only search 'level'
+ * directories deep. The parameter 'level' is the absolute maximum and is
+ * not related to restricts given to the '**' wildcard. If 'level' is 100
+ * and you use '**200' vim_findfile() will stop after 100 levels.
+ *
+ * 'filename' cannot contain wildcards!  It is used as-is, no backslashes to
+ * escape special characters.
+ *
+ * If 'stopdirs' is not NULL and nothing is found downward, the search is
+ * restarted on the next higher directory level. This is repeated until the
+ * start-directory of a search is contained in 'stopdirs'. 'stopdirs' has the
+ * format ";*<dirname>*\(;<dirname>\)*;\=$".
+ *
+ * If the 'path' is relative, the starting dir for the search is either VIM's
+ * current dir or if the path starts with "./" the current files dir.
+ * If the 'path' is absolute, the starting dir is that part of the path before
+ * the first wildcard.
+ *
+ * Upward search is only done on the starting dir.
+ *
+ * If 'free_visited' is TRUE the list of already visited files/directories is
+ * cleared. Set this to FALSE if you just want to search from another
+ * directory, but want to be sure that no directory from a previous search is
+ * searched again. This is useful if you search for a file at different places.
+ * The list of visited files/dirs can also be cleared with the function
+ * vim_findfile_free_visited().
+ *
+ * Set the parameter 'find_what' to FINDFILE_DIR if you want to search for
+ * directories only, FINDFILE_FILE for files only, FINDFILE_BOTH for both.
+ *
+ * A search context returned by a previous call to vim_findfile_init() can be
+ * passed in the parameter "search_ctx_arg".  This context is reused and
+ * reinitialized with the new parameters.  The list of already visited
+ * directories from this context is only deleted if the parameter
+ * "free_visited" is true.  Be aware that the passed "search_ctx_arg" is freed
+ * if the reinitialization fails.
+ *
+ * If you don't have a search context from a previous call "search_ctx_arg"
+ * must be NULL.
+ *
+ * This function silently ignores a few errors, vim_findfile() will have
+ * limited functionality then.
+ */
+    void *
+vim_findfile_init(
+    char_u	*path,
+    char_u	*filename,
+    char_u	*stopdirs UNUSED,
+    int		level,
+    int		free_visited,
+    int		find_what,
+    void	*search_ctx_arg,
+    int		tagfile,	// expanding names of tags files
+    char_u	*rel_fname)	// file name to use for "."
+{
+#ifdef FEAT_PATH_EXTRA
+    char_u		*wc_part;
+#endif
+    ff_stack_T		*sptr;
+    ff_search_ctx_T	*search_ctx;
+
+    // If a search context is given by the caller, reuse it, else allocate a
+    // new one.
+    if (search_ctx_arg != NULL)
+	search_ctx = search_ctx_arg;
+    else
+    {
+	search_ctx = ALLOC_CLEAR_ONE(ff_search_ctx_T);
+	if (search_ctx == NULL)
+	    goto error_return;
+    }
+    search_ctx->ffsc_find_what = find_what;
+    search_ctx->ffsc_tagfile = tagfile;
+
+    // clear the search context, but NOT the visited lists
+    ff_clear(search_ctx);
+
+    // clear visited list if wanted
+    if (free_visited == TRUE)
+	vim_findfile_free_visited(search_ctx);
+    else
+    {
+	// Reuse old visited lists. Get the visited list for the given
+	// filename. If no list for the current filename exists, creates a new
+	// one.
+	search_ctx->ffsc_visited_list = ff_get_visited_list(filename,
+					&search_ctx->ffsc_visited_lists_list);
+	if (search_ctx->ffsc_visited_list == NULL)
+	    goto error_return;
+	search_ctx->ffsc_dir_visited_list = ff_get_visited_list(filename,
+				    &search_ctx->ffsc_dir_visited_lists_list);
+	if (search_ctx->ffsc_dir_visited_list == NULL)
+	    goto error_return;
+    }
+
+    if (ff_expand_buffer == NULL)
+    {
+	ff_expand_buffer = alloc(MAXPATHL);
+	if (ff_expand_buffer == NULL)
+	    goto error_return;
+    }
+
+    // Store information on starting dir now if path is relative.
+    // If path is absolute, we do that later.
+    if (path[0] == '.'
+	    && (vim_ispathsep(path[1]) || path[1] == NUL)
+	    && (!tagfile || vim_strchr(p_cpo, CPO_DOTTAG) == NULL)
+	    && rel_fname != NULL)
+    {
+	int	len = (int)(gettail(rel_fname) - rel_fname);
+
+	if (!vim_isAbsName(rel_fname) && len + 1 < MAXPATHL)
+	{
+	    // Make the start dir an absolute path name.
+	    vim_strncpy(ff_expand_buffer, rel_fname, len);
+	    search_ctx->ffsc_start_dir = FullName_save(ff_expand_buffer, FALSE);
+	}
+	else
+	    search_ctx->ffsc_start_dir = vim_strnsave(rel_fname, len);
+	if (search_ctx->ffsc_start_dir == NULL)
+	    goto error_return;
+	if (*++path != NUL)
+	    ++path;
+    }
+    else if (*path == NUL || !vim_isAbsName(path))
+    {
+#ifdef BACKSLASH_IN_FILENAME
+	// "c:dir" needs "c:" to be expanded, otherwise use current dir
+	if (*path != NUL && path[1] == ':')
+	{
+	    char_u  drive[3];
+
+	    drive[0] = path[0];
+	    drive[1] = ':';
+	    drive[2] = NUL;
+	    if (vim_FullName(drive, ff_expand_buffer, MAXPATHL, TRUE) == FAIL)
+		goto error_return;
+	    path += 2;
+	}
+	else
+#endif
+	if (mch_dirname(ff_expand_buffer, MAXPATHL) == FAIL)
+	    goto error_return;
+
+	search_ctx->ffsc_start_dir = vim_strsave(ff_expand_buffer);
+	if (search_ctx->ffsc_start_dir == NULL)
+	    goto error_return;
+
+#ifdef BACKSLASH_IN_FILENAME
+	// A path that starts with "/dir" is relative to the drive, not to the
+	// directory (but not for "//machine/dir").  Only use the drive name.
+	if ((*path == '/' || *path == '\\')
+		&& path[1] != path[0]
+		&& search_ctx->ffsc_start_dir[1] == ':')
+	    search_ctx->ffsc_start_dir[2] = NUL;
+#endif
+    }
+
+#ifdef FEAT_PATH_EXTRA
+    /*
+     * If stopdirs are given, split them into an array of pointers.
+     * If this fails (mem allocation), there is no upward search at all or a
+     * stop directory is not recognized -> continue silently.
+     * If stopdirs just contains a ";" or is empty,
+     * search_ctx->ffsc_stopdirs_v will only contain a  NULL pointer. This
+     * is handled as unlimited upward search.  See function
+     * ff_path_in_stoplist() for details.
+     */
+    if (stopdirs != NULL)
+    {
+	char_u	*walker = stopdirs;
+	int	dircount;
+
+	while (*walker == ';')
+	    walker++;
+
+	dircount = 1;
+	search_ctx->ffsc_stopdirs_v = ALLOC_ONE(char_u *);
+
+	if (search_ctx->ffsc_stopdirs_v != NULL)
+	{
+	    do
+	    {
+		char_u	*helper;
+		void	*ptr;
+
+		helper = walker;
+		ptr = vim_realloc(search_ctx->ffsc_stopdirs_v,
+					   (dircount + 1) * sizeof(char_u *));
+		if (ptr)
+		    search_ctx->ffsc_stopdirs_v = ptr;
+		else
+		    // ignore, keep what we have and continue
+		    break;
+		walker = vim_strchr(walker, ';');
+		if (walker)
+		{
+		    search_ctx->ffsc_stopdirs_v[dircount-1] =
+					 vim_strnsave(helper, walker - helper);
+		    walker++;
+		}
+		else
+		    // this might be "", which means ascent till top
+		    // of directory tree.
+		    search_ctx->ffsc_stopdirs_v[dircount-1] =
+							  vim_strsave(helper);
+
+		dircount++;
+
+	    } while (walker != NULL);
+	    search_ctx->ffsc_stopdirs_v[dircount-1] = NULL;
+	}
+    }
+#endif
+
+#ifdef FEAT_PATH_EXTRA
+    search_ctx->ffsc_level = level;
+
+    /*
+     * split into:
+     *  -fix path
+     *  -wildcard_stuff (might be NULL)
+     */
+    wc_part = vim_strchr(path, '*');
+    if (wc_part != NULL)
+    {
+	int	llevel;
+	int	len;
+	char	*errpt;
+
+	// save the fix part of the path
+	search_ctx->ffsc_fix_path = vim_strnsave(path, wc_part - path);
+
+	/*
+	 * copy wc_path and add restricts to the '**' wildcard.
+	 * The octet after a '**' is used as a (binary) counter.
+	 * So '**3' is transposed to '**^C' ('^C' is ASCII value 3)
+	 * or '**76' is transposed to '**N'( 'N' is ASCII value 76).
+	 * For EBCDIC you get different character values.
+	 * If no restrict is given after '**' the default is used.
+	 * Due to this technique the path looks awful if you print it as a
+	 * string.
+	 */
+	len = 0;
+	while (*wc_part != NUL)
+	{
+	    if (len + 5 >= MAXPATHL)
+	    {
+		emsg(_(e_pathtoolong));
+		break;
+	    }
+	    if (STRNCMP(wc_part, "**", 2) == 0)
+	    {
+		ff_expand_buffer[len++] = *wc_part++;
+		ff_expand_buffer[len++] = *wc_part++;
+
+		llevel = strtol((char *)wc_part, &errpt, 10);
+		if ((char_u *)errpt != wc_part && llevel > 0 && llevel < 255)
+		    ff_expand_buffer[len++] = llevel;
+		else if ((char_u *)errpt != wc_part && llevel == 0)
+		    // restrict is 0 -> remove already added '**'
+		    len -= 2;
+		else
+		    ff_expand_buffer[len++] = FF_MAX_STAR_STAR_EXPAND;
+		wc_part = (char_u *)errpt;
+		if (*wc_part != NUL && !vim_ispathsep(*wc_part))
+		{
+		    semsg(_("E343: Invalid path: '**[number]' must be at the end of the path or be followed by '%s'."), PATHSEPSTR);
+		    goto error_return;
+		}
+	    }
+	    else
+		ff_expand_buffer[len++] = *wc_part++;
+	}
+	ff_expand_buffer[len] = NUL;
+	search_ctx->ffsc_wc_path = vim_strsave(ff_expand_buffer);
+
+	if (search_ctx->ffsc_wc_path == NULL)
+	    goto error_return;
+    }
+    else
+#endif
+	search_ctx->ffsc_fix_path = vim_strsave(path);
+
+    if (search_ctx->ffsc_start_dir == NULL)
+    {
+	// store the fix part as startdir.
+	// This is needed if the parameter path is fully qualified.
+	search_ctx->ffsc_start_dir = vim_strsave(search_ctx->ffsc_fix_path);
+	if (search_ctx->ffsc_start_dir == NULL)
+	    goto error_return;
+	search_ctx->ffsc_fix_path[0] = NUL;
+    }
+
+    // create an absolute path
+    if (STRLEN(search_ctx->ffsc_start_dir)
+			  + STRLEN(search_ctx->ffsc_fix_path) + 3 >= MAXPATHL)
+    {
+	emsg(_(e_pathtoolong));
+	goto error_return;
+    }
+    STRCPY(ff_expand_buffer, search_ctx->ffsc_start_dir);
+    add_pathsep(ff_expand_buffer);
+    {
+	int    eb_len = (int)STRLEN(ff_expand_buffer);
+	char_u *buf = alloc(eb_len
+				+ (int)STRLEN(search_ctx->ffsc_fix_path) + 1);
+
+	STRCPY(buf, ff_expand_buffer);
+	STRCPY(buf + eb_len, search_ctx->ffsc_fix_path);
+	if (mch_isdir(buf))
+	{
+	    STRCAT(ff_expand_buffer, search_ctx->ffsc_fix_path);
+	    add_pathsep(ff_expand_buffer);
+	}
+#ifdef FEAT_PATH_EXTRA
+	else
+	{
+	    char_u *p =  gettail(search_ctx->ffsc_fix_path);
+	    char_u *wc_path = NULL;
+	    char_u *temp = NULL;
+	    int    len = 0;
+
+	    if (p > search_ctx->ffsc_fix_path)
+	    {
+		len = (int)(p - search_ctx->ffsc_fix_path) - 1;
+		STRNCAT(ff_expand_buffer, search_ctx->ffsc_fix_path, len);
+		add_pathsep(ff_expand_buffer);
+	    }
+	    else
+		len = (int)STRLEN(search_ctx->ffsc_fix_path);
+
+	    if (search_ctx->ffsc_wc_path != NULL)
+	    {
+		wc_path = vim_strsave(search_ctx->ffsc_wc_path);
+		temp = alloc(STRLEN(search_ctx->ffsc_wc_path)
+				 + STRLEN(search_ctx->ffsc_fix_path + len)
+				 + 1);
+		if (temp == NULL || wc_path == NULL)
+		{
+		    vim_free(buf);
+		    vim_free(temp);
+		    vim_free(wc_path);
+		    goto error_return;
+		}
+
+		STRCPY(temp, search_ctx->ffsc_fix_path + len);
+		STRCAT(temp, search_ctx->ffsc_wc_path);
+		vim_free(search_ctx->ffsc_wc_path);
+		vim_free(wc_path);
+		search_ctx->ffsc_wc_path = temp;
+	    }
+	}
+#endif
+	vim_free(buf);
+    }
+
+    sptr = ff_create_stack_element(ff_expand_buffer,
+#ifdef FEAT_PATH_EXTRA
+	    search_ctx->ffsc_wc_path,
+#endif
+	    level, 0);
+
+    if (sptr == NULL)
+	goto error_return;
+
+    ff_push(search_ctx, sptr);
+
+    search_ctx->ffsc_file_to_search = vim_strsave(filename);
+    if (search_ctx->ffsc_file_to_search == NULL)
+	goto error_return;
+
+    return search_ctx;
+
+error_return:
+    /*
+     * We clear the search context now!
+     * Even when the caller gave us a (perhaps valid) context we free it here,
+     * as we might have already destroyed it.
+     */
+    vim_findfile_cleanup(search_ctx);
+    return NULL;
+}
+
+#if defined(FEAT_PATH_EXTRA) || defined(PROTO)
+/*
+ * Get the stopdir string.  Check that ';' is not escaped.
+ */
+    char_u *
+vim_findfile_stopdir(char_u *buf)
+{
+    char_u	*r_ptr = buf;
+
+    while (*r_ptr != NUL && *r_ptr != ';')
+    {
+	if (r_ptr[0] == '\\' && r_ptr[1] == ';')
+	{
+	    // Overwrite the escape char,
+	    // use STRLEN(r_ptr) to move the trailing '\0'.
+	    STRMOVE(r_ptr, r_ptr + 1);
+	    r_ptr++;
+	}
+	r_ptr++;
+    }
+    if (*r_ptr == ';')
+    {
+	*r_ptr = 0;
+	r_ptr++;
+    }
+    else if (*r_ptr == NUL)
+	r_ptr = NULL;
+    return r_ptr;
+}
+#endif
+
+/*
+ * Clean up the given search context. Can handle a NULL pointer.
+ */
+    void
+vim_findfile_cleanup(void *ctx)
+{
+    if (ctx == NULL)
+	return;
+
+    vim_findfile_free_visited(ctx);
+    ff_clear(ctx);
+    vim_free(ctx);
+}
+
+/*
+ * Find a file in a search context.
+ * The search context was created with vim_findfile_init() above.
+ * Return a pointer to an allocated file name or NULL if nothing found.
+ * To get all matching files call this function until you get NULL.
+ *
+ * If the passed search_context is NULL, NULL is returned.
+ *
+ * The search algorithm is depth first. To change this replace the
+ * stack with a list (don't forget to leave partly searched directories on the
+ * top of the list).
+ */
+    char_u *
+vim_findfile(void *search_ctx_arg)
+{
+    char_u	*file_path;
+#ifdef FEAT_PATH_EXTRA
+    char_u	*rest_of_wildcards;
+    char_u	*path_end = NULL;
+#endif
+    ff_stack_T	*stackp;
+#if defined(FEAT_SEARCHPATH) || defined(FEAT_PATH_EXTRA)
+    int		len;
+#endif
+    int		i;
+    char_u	*p;
+#ifdef FEAT_SEARCHPATH
+    char_u	*suf;
+#endif
+    ff_search_ctx_T *search_ctx;
+
+    if (search_ctx_arg == NULL)
+	return NULL;
+
+    search_ctx = (ff_search_ctx_T *)search_ctx_arg;
+
+    /*
+     * filepath is used as buffer for various actions and as the storage to
+     * return a found filename.
+     */
+    if ((file_path = alloc(MAXPATHL)) == NULL)
+	return NULL;
+
+#ifdef FEAT_PATH_EXTRA
+    // store the end of the start dir -- needed for upward search
+    if (search_ctx->ffsc_start_dir != NULL)
+	path_end = &search_ctx->ffsc_start_dir[
+					  STRLEN(search_ctx->ffsc_start_dir)];
+#endif
+
+#ifdef FEAT_PATH_EXTRA
+    // upward search loop
+    for (;;)
+    {
+#endif
+	// downward search loop
+	for (;;)
+	{
+	    // check if user user wants to stop the search
+	    ui_breakcheck();
+	    if (got_int)
+		break;
+
+	    // get directory to work on from stack
+	    stackp = ff_pop(search_ctx);
+	    if (stackp == NULL)
+		break;
+
+	    /*
+	     * TODO: decide if we leave this test in
+	     *
+	     * GOOD: don't search a directory(-tree) twice.
+	     * BAD:  - check linked list for every new directory entered.
+	     *       - check for double files also done below
+	     *
+	     * Here we check if we already searched this directory.
+	     * We already searched a directory if:
+	     * 1) The directory is the same.
+	     * 2) We would use the same wildcard string.
+	     *
+	     * Good if you have links on same directory via several ways
+	     *  or you have selfreferences in directories (e.g. SuSE Linux 6.3:
+	     *  /etc/rc.d/init.d is linked to /etc/rc.d -> endless loop)
+	     *
+	     * This check is only needed for directories we work on for the
+	     * first time (hence stackp->ff_filearray == NULL)
+	     */
+	    if (stackp->ffs_filearray == NULL
+		    && ff_check_visited(&search_ctx->ffsc_dir_visited_list
+							  ->ffvl_visited_list,
+			stackp->ffs_fix_path
+#ifdef FEAT_PATH_EXTRA
+			, stackp->ffs_wc_path
+#endif
+			) == FAIL)
+	    {
+#ifdef FF_VERBOSE
+		if (p_verbose >= 5)
+		{
+		    verbose_enter_scroll();
+		    smsg("Already Searched: %s (%s)",
+				   stackp->ffs_fix_path, stackp->ffs_wc_path);
+		    // don't overwrite this either
+		    msg_puts("\n");
+		    verbose_leave_scroll();
+		}
+#endif
+		ff_free_stack_element(stackp);
+		continue;
+	    }
+#ifdef FF_VERBOSE
+	    else if (p_verbose >= 5)
+	    {
+		verbose_enter_scroll();
+		smsg("Searching: %s (%s)",
+				   stackp->ffs_fix_path, stackp->ffs_wc_path);
+		// don't overwrite this either
+		msg_puts("\n");
+		verbose_leave_scroll();
+	    }
+#endif
+
+	    // check depth
+	    if (stackp->ffs_level <= 0)
+	    {
+		ff_free_stack_element(stackp);
+		continue;
+	    }
+
+	    file_path[0] = NUL;
+
+	    /*
+	     * If no filearray till now expand wildcards
+	     * The function expand_wildcards() can handle an array of paths
+	     * and all possible expands are returned in one array. We use this
+	     * to handle the expansion of '**' into an empty string.
+	     */
+	    if (stackp->ffs_filearray == NULL)
+	    {
+		char_u *dirptrs[2];
+
+		// we use filepath to build the path expand_wildcards() should
+		// expand.
+		dirptrs[0] = file_path;
+		dirptrs[1] = NULL;
+
+		// if we have a start dir copy it in
+		if (!vim_isAbsName(stackp->ffs_fix_path)
+						&& search_ctx->ffsc_start_dir)
+		{
+		    if (STRLEN(search_ctx->ffsc_start_dir) + 1 < MAXPATHL)
+		    {
+			STRCPY(file_path, search_ctx->ffsc_start_dir);
+			add_pathsep(file_path);
+		    }
+		    else
+		    {
+			ff_free_stack_element(stackp);
+			goto fail;
+		    }
+		}
+
+		// append the fix part of the search path
+		if (STRLEN(file_path) + STRLEN(stackp->ffs_fix_path) + 1
+								    < MAXPATHL)
+		{
+		    STRCAT(file_path, stackp->ffs_fix_path);
+		    add_pathsep(file_path);
+		}
+		else
+		{
+		    ff_free_stack_element(stackp);
+		    goto fail;
+		}
+
+#ifdef FEAT_PATH_EXTRA
+		rest_of_wildcards = stackp->ffs_wc_path;
+		if (*rest_of_wildcards != NUL)
+		{
+		    len = (int)STRLEN(file_path);
+		    if (STRNCMP(rest_of_wildcards, "**", 2) == 0)
+		    {
+			// pointer to the restrict byte
+			// The restrict byte is not a character!
+			p = rest_of_wildcards + 2;
+
+			if (*p > 0)
+			{
+			    (*p)--;
+			    if (len + 1 < MAXPATHL)
+				file_path[len++] = '*';
+			    else
+			    {
+				ff_free_stack_element(stackp);
+				goto fail;
+			    }
+			}
+
+			if (*p == 0)
+			{
+			    // remove '**<numb> from wildcards
+			    STRMOVE(rest_of_wildcards, rest_of_wildcards + 3);
+			}
+			else
+			    rest_of_wildcards += 3;
+
+			if (stackp->ffs_star_star_empty == 0)
+			{
+			    // if not done before, expand '**' to empty
+			    stackp->ffs_star_star_empty = 1;
+			    dirptrs[1] = stackp->ffs_fix_path;
+			}
+		    }
+
+		    /*
+		     * Here we copy until the next path separator or the end of
+		     * the path. If we stop at a path separator, there is
+		     * still something else left. This is handled below by
+		     * pushing every directory returned from expand_wildcards()
+		     * on the stack again for further search.
+		     */
+		    while (*rest_of_wildcards
+			    && !vim_ispathsep(*rest_of_wildcards))
+			if (len + 1 < MAXPATHL)
+			    file_path[len++] = *rest_of_wildcards++;
+			else
+			{
+			    ff_free_stack_element(stackp);
+			    goto fail;
+			}
+
+		    file_path[len] = NUL;
+		    if (vim_ispathsep(*rest_of_wildcards))
+			rest_of_wildcards++;
+		}
+#endif
+
+		/*
+		 * Expand wildcards like "*" and "$VAR".
+		 * If the path is a URL don't try this.
+		 */
+		if (path_with_url(dirptrs[0]))
+		{
+		    stackp->ffs_filearray = ALLOC_ONE(char_u *);
+		    if (stackp->ffs_filearray != NULL
+			    && (stackp->ffs_filearray[0]
+				= vim_strsave(dirptrs[0])) != NULL)
+			stackp->ffs_filearray_size = 1;
+		    else
+			stackp->ffs_filearray_size = 0;
+		}
+		else
+		    // Add EW_NOTWILD because the expanded path may contain
+		    // wildcard characters that are to be taken literally.
+		    // This is a bit of a hack.
+		    expand_wildcards((dirptrs[1] == NULL) ? 1 : 2, dirptrs,
+			    &stackp->ffs_filearray_size,
+			    &stackp->ffs_filearray,
+			    EW_DIR|EW_ADDSLASH|EW_SILENT|EW_NOTWILD);
+
+		stackp->ffs_filearray_cur = 0;
+		stackp->ffs_stage = 0;
+	    }
+#ifdef FEAT_PATH_EXTRA
+	    else
+		rest_of_wildcards = &stackp->ffs_wc_path[
+						 STRLEN(stackp->ffs_wc_path)];
+#endif
+
+	    if (stackp->ffs_stage == 0)
+	    {
+		// this is the first time we work on this directory
+#ifdef FEAT_PATH_EXTRA
+		if (*rest_of_wildcards == NUL)
+#endif
+		{
+		    /*
+		     * We don't have further wildcards to expand, so we have to
+		     * check for the final file now.
+		     */
+		    for (i = stackp->ffs_filearray_cur;
+					  i < stackp->ffs_filearray_size; ++i)
+		    {
+			if (!path_with_url(stackp->ffs_filearray[i])
+				      && !mch_isdir(stackp->ffs_filearray[i]))
+			    continue;   // not a directory
+
+			// prepare the filename to be checked for existence
+			// below
+			if (STRLEN(stackp->ffs_filearray[i]) + 1
+				+ STRLEN(search_ctx->ffsc_file_to_search)
+								    < MAXPATHL)
+			{
+			    STRCPY(file_path, stackp->ffs_filearray[i]);
+			    add_pathsep(file_path);
+			    STRCAT(file_path, search_ctx->ffsc_file_to_search);
+			}
+			else
+			{
+			    ff_free_stack_element(stackp);
+			    goto fail;
+			}
+
+			/*
+			 * Try without extra suffix and then with suffixes
+			 * from 'suffixesadd'.
+			 */
+#ifdef FEAT_SEARCHPATH
+			len = (int)STRLEN(file_path);
+			if (search_ctx->ffsc_tagfile)
+			    suf = (char_u *)"";
+			else
+			    suf = curbuf->b_p_sua;
+			for (;;)
+#endif
+			{
+			    // if file exists and we didn't already find it
+			    if ((path_with_url(file_path)
+				  || (mch_getperm(file_path) >= 0
+				      && (search_ctx->ffsc_find_what
+							      == FINDFILE_BOTH
+					  || ((search_ctx->ffsc_find_what
+							      == FINDFILE_DIR)
+						   == mch_isdir(file_path)))))
+#ifndef FF_VERBOSE
+				    && (ff_check_visited(
+					    &search_ctx->ffsc_visited_list->ffvl_visited_list,
+					    file_path
+#ifdef FEAT_PATH_EXTRA
+					    , (char_u *)""
+#endif
+					    ) == OK)
+#endif
+			       )
+			    {
+#ifdef FF_VERBOSE
+				if (ff_check_visited(
+					    &search_ctx->ffsc_visited_list->ffvl_visited_list,
+					    file_path
+#ifdef FEAT_PATH_EXTRA
+					    , (char_u *)""
+#endif
+						    ) == FAIL)
+				{
+				    if (p_verbose >= 5)
+				    {
+					verbose_enter_scroll();
+					smsg("Already: %s",
+								   file_path);
+					// don't overwrite this either
+					msg_puts("\n");
+					verbose_leave_scroll();
+				    }
+				    continue;
+				}
+#endif
+
+				// push dir to examine rest of subdirs later
+				stackp->ffs_filearray_cur = i + 1;
+				ff_push(search_ctx, stackp);
+
+				if (!path_with_url(file_path))
+				    simplify_filename(file_path);
+				if (mch_dirname(ff_expand_buffer, MAXPATHL)
+									== OK)
+				{
+				    p = shorten_fname(file_path,
+							    ff_expand_buffer);
+				    if (p != NULL)
+					STRMOVE(file_path, p);
+				}
+#ifdef FF_VERBOSE
+				if (p_verbose >= 5)
+				{
+				    verbose_enter_scroll();
+				    smsg("HIT: %s", file_path);
+				    // don't overwrite this either
+				    msg_puts("\n");
+				    verbose_leave_scroll();
+				}
+#endif
+				return file_path;
+			    }
+
+#ifdef FEAT_SEARCHPATH
+			    // Not found or found already, try next suffix.
+			    if (*suf == NUL)
+				break;
+			    copy_option_part(&suf, file_path + len,
+							 MAXPATHL - len, ",");
+#endif
+			}
+		    }
+		}
+#ifdef FEAT_PATH_EXTRA
+		else
+		{
+		    /*
+		     * still wildcards left, push the directories for further
+		     * search
+		     */
+		    for (i = stackp->ffs_filearray_cur;
+					  i < stackp->ffs_filearray_size; ++i)
+		    {
+			if (!mch_isdir(stackp->ffs_filearray[i]))
+			    continue;	// not a directory
+
+			ff_push(search_ctx,
+				ff_create_stack_element(
+						     stackp->ffs_filearray[i],
+						     rest_of_wildcards,
+						     stackp->ffs_level - 1, 0));
+		    }
+		}
+#endif
+		stackp->ffs_filearray_cur = 0;
+		stackp->ffs_stage = 1;
+	    }
+
+#ifdef FEAT_PATH_EXTRA
+	    /*
+	     * if wildcards contains '**' we have to descent till we reach the
+	     * leaves of the directory tree.
+	     */
+	    if (STRNCMP(stackp->ffs_wc_path, "**", 2) == 0)
+	    {
+		for (i = stackp->ffs_filearray_cur;
+					  i < stackp->ffs_filearray_size; ++i)
+		{
+		    if (fnamecmp(stackp->ffs_filearray[i],
+						   stackp->ffs_fix_path) == 0)
+			continue; // don't repush same directory
+		    if (!mch_isdir(stackp->ffs_filearray[i]))
+			continue;   // not a directory
+		    ff_push(search_ctx,
+			    ff_create_stack_element(stackp->ffs_filearray[i],
+				stackp->ffs_wc_path, stackp->ffs_level - 1, 1));
+		}
+	    }
+#endif
+
+	    // we are done with the current directory
+	    ff_free_stack_element(stackp);
+
+	}
+
+#ifdef FEAT_PATH_EXTRA
+	// If we reached this, we didn't find anything downwards.
+	// Let's check if we should do an upward search.
+	if (search_ctx->ffsc_start_dir
+		&& search_ctx->ffsc_stopdirs_v != NULL && !got_int)
+	{
+	    ff_stack_T  *sptr;
+
+	    // is the last starting directory in the stop list?
+	    if (ff_path_in_stoplist(search_ctx->ffsc_start_dir,
+		       (int)(path_end - search_ctx->ffsc_start_dir),
+		       search_ctx->ffsc_stopdirs_v) == TRUE)
+		break;
+
+	    // cut of last dir
+	    while (path_end > search_ctx->ffsc_start_dir
+						  && vim_ispathsep(*path_end))
+		path_end--;
+	    while (path_end > search_ctx->ffsc_start_dir
+					      && !vim_ispathsep(path_end[-1]))
+		path_end--;
+	    *path_end = 0;
+	    path_end--;
+
+	    if (*search_ctx->ffsc_start_dir == 0)
+		break;
+
+	    if (STRLEN(search_ctx->ffsc_start_dir) + 1
+		    + STRLEN(search_ctx->ffsc_fix_path) < MAXPATHL)
+	    {
+		STRCPY(file_path, search_ctx->ffsc_start_dir);
+		add_pathsep(file_path);
+		STRCAT(file_path, search_ctx->ffsc_fix_path);
+	    }
+	    else
+		goto fail;
+
+	    // create a new stack entry
+	    sptr = ff_create_stack_element(file_path,
+		    search_ctx->ffsc_wc_path, search_ctx->ffsc_level, 0);
+	    if (sptr == NULL)
+		break;
+	    ff_push(search_ctx, sptr);
+	}
+	else
+	    break;
+    }
+#endif
+
+fail:
+    vim_free(file_path);
+    return NULL;
+}
+
+/*
+ * Free the list of lists of visited files and directories
+ * Can handle it if the passed search_context is NULL;
+ */
+    static void
+vim_findfile_free_visited(void *search_ctx_arg)
+{
+    ff_search_ctx_T *search_ctx;
+
+    if (search_ctx_arg == NULL)
+	return;
+
+    search_ctx = (ff_search_ctx_T *)search_ctx_arg;
+    vim_findfile_free_visited_list(&search_ctx->ffsc_visited_lists_list);
+    vim_findfile_free_visited_list(&search_ctx->ffsc_dir_visited_lists_list);
+}
+
+    static void
+vim_findfile_free_visited_list(ff_visited_list_hdr_T **list_headp)
+{
+    ff_visited_list_hdr_T *vp;
+
+    while (*list_headp != NULL)
+    {
+	vp = (*list_headp)->ffvl_next;
+	ff_free_visited_list((*list_headp)->ffvl_visited_list);
+
+	vim_free((*list_headp)->ffvl_filename);
+	vim_free(*list_headp);
+	*list_headp = vp;
+    }
+    *list_headp = NULL;
+}
+
+    static void
+ff_free_visited_list(ff_visited_T *vl)
+{
+    ff_visited_T *vp;
+
+    while (vl != NULL)
+    {
+	vp = vl->ffv_next;
+#ifdef FEAT_PATH_EXTRA
+	vim_free(vl->ffv_wc_path);
+#endif
+	vim_free(vl);
+	vl = vp;
+    }
+    vl = NULL;
+}
+
+/*
+ * Returns the already visited list for the given filename. If none is found it
+ * allocates a new one.
+ */
+    static ff_visited_list_hdr_T*
+ff_get_visited_list(
+    char_u			*filename,
+    ff_visited_list_hdr_T	**list_headp)
+{
+    ff_visited_list_hdr_T  *retptr = NULL;
+
+    // check if a visited list for the given filename exists
+    if (*list_headp != NULL)
+    {
+	retptr = *list_headp;
+	while (retptr != NULL)
+	{
+	    if (fnamecmp(filename, retptr->ffvl_filename) == 0)
+	    {
+#ifdef FF_VERBOSE
+		if (p_verbose >= 5)
+		{
+		    verbose_enter_scroll();
+		    smsg("ff_get_visited_list: FOUND list for %s",
+								    filename);
+		    // don't overwrite this either
+		    msg_puts("\n");
+		    verbose_leave_scroll();
+		}
+#endif
+		return retptr;
+	    }
+	    retptr = retptr->ffvl_next;
+	}
+    }
+
+#ifdef FF_VERBOSE
+    if (p_verbose >= 5)
+    {
+	verbose_enter_scroll();
+	smsg("ff_get_visited_list: new list for %s", filename);
+	// don't overwrite this either
+	msg_puts("\n");
+	verbose_leave_scroll();
+    }
+#endif
+
+    /*
+     * if we reach this we didn't find a list and we have to allocate new list
+     */
+    retptr = ALLOC_ONE(ff_visited_list_hdr_T);
+    if (retptr == NULL)
+	return NULL;
+
+    retptr->ffvl_visited_list = NULL;
+    retptr->ffvl_filename = vim_strsave(filename);
+    if (retptr->ffvl_filename == NULL)
+    {
+	vim_free(retptr);
+	return NULL;
+    }
+    retptr->ffvl_next = *list_headp;
+    *list_headp = retptr;
+
+    return retptr;
+}
+
+#ifdef FEAT_PATH_EXTRA
+/*
+ * check if two wildcard paths are equal. Returns TRUE or FALSE.
+ * They are equal if:
+ *  - both paths are NULL
+ *  - they have the same length
+ *  - char by char comparison is OK
+ *  - the only differences are in the counters behind a '**', so
+ *    '**\20' is equal to '**\24'
+ */
+    static int
+ff_wc_equal(char_u *s1, char_u *s2)
+{
+    int		i, j;
+    int		c1 = NUL;
+    int		c2 = NUL;
+    int		prev1 = NUL;
+    int		prev2 = NUL;
+
+    if (s1 == s2)
+	return TRUE;
+
+    if (s1 == NULL || s2 == NULL)
+	return FALSE;
+
+    for (i = 0, j = 0; s1[i] != NUL && s2[j] != NUL;)
+    {
+	c1 = PTR2CHAR(s1 + i);
+	c2 = PTR2CHAR(s2 + j);
+
+	if ((p_fic ? MB_TOLOWER(c1) != MB_TOLOWER(c2) : c1 != c2)
+		&& (prev1 != '*' || prev2 != '*'))
+	    return FALSE;
+	prev2 = prev1;
+	prev1 = c1;
+
+	i += mb_ptr2len(s1 + i);
+	j += mb_ptr2len(s2 + j);
+    }
+    return s1[i] == s2[j];
+}
+#endif
+
+/*
+ * maintains the list of already visited files and dirs
+ * returns FAIL if the given file/dir is already in the list
+ * returns OK if it is newly added
+ *
+ * TODO: What to do on memory allocation problems?
+ *	 -> return TRUE - Better the file is found several times instead of
+ *	    never.
+ */
+    static int
+ff_check_visited(
+    ff_visited_T	**visited_list,
+    char_u		*fname
+#ifdef FEAT_PATH_EXTRA
+    , char_u		*wc_path
+#endif
+    )
+{
+    ff_visited_T	*vp;
+#ifdef UNIX
+    stat_T		st;
+    int			url = FALSE;
+#endif
+
+    // For an URL we only compare the name, otherwise we compare the
+    // device/inode (unix) or the full path name (not Unix).
+    if (path_with_url(fname))
+    {
+	vim_strncpy(ff_expand_buffer, fname, MAXPATHL - 1);
+#ifdef UNIX
+	url = TRUE;
+#endif
+    }
+    else
+    {
+	ff_expand_buffer[0] = NUL;
+#ifdef UNIX
+	if (mch_stat((char *)fname, &st) < 0)
+#else
+	if (vim_FullName(fname, ff_expand_buffer, MAXPATHL, TRUE) == FAIL)
+#endif
+	    return FAIL;
+    }
+
+    // check against list of already visited files
+    for (vp = *visited_list; vp != NULL; vp = vp->ffv_next)
+    {
+	if (
+#ifdef UNIX
+		!url ? (vp->ffv_dev_valid && vp->ffv_dev == st.st_dev
+						  && vp->ffv_ino == st.st_ino)
+		     :
+#endif
+		fnamecmp(vp->ffv_fname, ff_expand_buffer) == 0
+	   )
+	{
+#ifdef FEAT_PATH_EXTRA
+	    // are the wildcard parts equal
+	    if (ff_wc_equal(vp->ffv_wc_path, wc_path) == TRUE)
+#endif
+		// already visited
+		return FAIL;
+	}
+    }
+
+    /*
+     * New file/dir.  Add it to the list of visited files/dirs.
+     */
+    vp = alloc(sizeof(ff_visited_T) + STRLEN(ff_expand_buffer));
+
+    if (vp != NULL)
+    {
+#ifdef UNIX
+	if (!url)
+	{
+	    vp->ffv_dev_valid = TRUE;
+	    vp->ffv_ino = st.st_ino;
+	    vp->ffv_dev = st.st_dev;
+	    vp->ffv_fname[0] = NUL;
+	}
+	else
+	{
+	    vp->ffv_dev_valid = FALSE;
+#endif
+	    STRCPY(vp->ffv_fname, ff_expand_buffer);
+#ifdef UNIX
+	}
+#endif
+#ifdef FEAT_PATH_EXTRA
+	if (wc_path != NULL)
+	    vp->ffv_wc_path = vim_strsave(wc_path);
+	else
+	    vp->ffv_wc_path = NULL;
+#endif
+
+	vp->ffv_next = *visited_list;
+	*visited_list = vp;
+    }
+
+    return OK;
+}
+
+/*
+ * create stack element from given path pieces
+ */
+    static ff_stack_T *
+ff_create_stack_element(
+    char_u	*fix_part,
+#ifdef FEAT_PATH_EXTRA
+    char_u	*wc_part,
+#endif
+    int		level,
+    int		star_star_empty)
+{
+    ff_stack_T	*new;
+
+    new = ALLOC_ONE(ff_stack_T);
+    if (new == NULL)
+	return NULL;
+
+    new->ffs_prev	   = NULL;
+    new->ffs_filearray	   = NULL;
+    new->ffs_filearray_size = 0;
+    new->ffs_filearray_cur  = 0;
+    new->ffs_stage	   = 0;
+    new->ffs_level	   = level;
+    new->ffs_star_star_empty = star_star_empty;
+
+    // the following saves NULL pointer checks in vim_findfile
+    if (fix_part == NULL)
+	fix_part = (char_u *)"";
+    new->ffs_fix_path = vim_strsave(fix_part);
+
+#ifdef FEAT_PATH_EXTRA
+    if (wc_part == NULL)
+	wc_part  = (char_u *)"";
+    new->ffs_wc_path = vim_strsave(wc_part);
+#endif
+
+    if (new->ffs_fix_path == NULL
+#ifdef FEAT_PATH_EXTRA
+	    || new->ffs_wc_path == NULL
+#endif
+	    )
+    {
+	ff_free_stack_element(new);
+	new = NULL;
+    }
+
+    return new;
+}
+
+/*
+ * Push a dir on the directory stack.
+ */
+    static void
+ff_push(ff_search_ctx_T *search_ctx, ff_stack_T *stack_ptr)
+{
+    // check for NULL pointer, not to return an error to the user, but
+    // to prevent a crash
+    if (stack_ptr != NULL)
+    {
+	stack_ptr->ffs_prev = search_ctx->ffsc_stack_ptr;
+	search_ctx->ffsc_stack_ptr = stack_ptr;
+    }
+}
+
+/*
+ * Pop a dir from the directory stack.
+ * Returns NULL if stack is empty.
+ */
+    static ff_stack_T *
+ff_pop(ff_search_ctx_T *search_ctx)
+{
+    ff_stack_T  *sptr;
+
+    sptr = search_ctx->ffsc_stack_ptr;
+    if (search_ctx->ffsc_stack_ptr != NULL)
+	search_ctx->ffsc_stack_ptr = search_ctx->ffsc_stack_ptr->ffs_prev;
+
+    return sptr;
+}
+
+/*
+ * free the given stack element
+ */
+    static void
+ff_free_stack_element(ff_stack_T *stack_ptr)
+{
+    // vim_free handles possible NULL pointers
+    vim_free(stack_ptr->ffs_fix_path);
+#ifdef FEAT_PATH_EXTRA
+    vim_free(stack_ptr->ffs_wc_path);
+#endif
+
+    if (stack_ptr->ffs_filearray != NULL)
+	FreeWild(stack_ptr->ffs_filearray_size, stack_ptr->ffs_filearray);
+
+    vim_free(stack_ptr);
+}
+
+/*
+ * Clear the search context, but NOT the visited list.
+ */
+    static void
+ff_clear(ff_search_ctx_T *search_ctx)
+{
+    ff_stack_T   *sptr;
+
+    // clear up stack
+    while ((sptr = ff_pop(search_ctx)) != NULL)
+	ff_free_stack_element(sptr);
+
+    vim_free(search_ctx->ffsc_file_to_search);
+    vim_free(search_ctx->ffsc_start_dir);
+    vim_free(search_ctx->ffsc_fix_path);
+#ifdef FEAT_PATH_EXTRA
+    vim_free(search_ctx->ffsc_wc_path);
+#endif
+
+#ifdef FEAT_PATH_EXTRA
+    if (search_ctx->ffsc_stopdirs_v != NULL)
+    {
+	int  i = 0;
+
+	while (search_ctx->ffsc_stopdirs_v[i] != NULL)
+	{
+	    vim_free(search_ctx->ffsc_stopdirs_v[i]);
+	    i++;
+	}
+	vim_free(search_ctx->ffsc_stopdirs_v);
+    }
+    search_ctx->ffsc_stopdirs_v = NULL;
+#endif
+
+    // reset everything
+    search_ctx->ffsc_file_to_search = NULL;
+    search_ctx->ffsc_start_dir = NULL;
+    search_ctx->ffsc_fix_path = NULL;
+#ifdef FEAT_PATH_EXTRA
+    search_ctx->ffsc_wc_path = NULL;
+    search_ctx->ffsc_level = 0;
+#endif
+}
+
+#ifdef FEAT_PATH_EXTRA
+/*
+ * check if the given path is in the stopdirs
+ * returns TRUE if yes else FALSE
+ */
+    static int
+ff_path_in_stoplist(char_u *path, int path_len, char_u **stopdirs_v)
+{
+    int		i = 0;
+
+    // eat up trailing path separators, except the first
+    while (path_len > 1 && vim_ispathsep(path[path_len - 1]))
+	path_len--;
+
+    // if no path consider it as match
+    if (path_len == 0)
+	return TRUE;
+
+    for (i = 0; stopdirs_v[i] != NULL; i++)
+    {
+	if ((int)STRLEN(stopdirs_v[i]) > path_len)
+	{
+	    // match for parent directory. So '/home' also matches
+	    // '/home/rks'. Check for PATHSEP in stopdirs_v[i], else
+	    // '/home/r' would also match '/home/rks'
+	    if (fnamencmp(stopdirs_v[i], path, path_len) == 0
+		    && vim_ispathsep(stopdirs_v[i][path_len]))
+		return TRUE;
+	}
+	else
+	{
+	    if (fnamecmp(stopdirs_v[i], path) == 0)
+		return TRUE;
+	}
+    }
+    return FALSE;
+}
+#endif
+
+#if defined(FEAT_SEARCHPATH) || defined(PROTO)
+/*
+ * Find the file name "ptr[len]" in the path.  Also finds directory names.
+ *
+ * On the first call set the parameter 'first' to TRUE to initialize
+ * the search.  For repeating calls to FALSE.
+ *
+ * Repeating calls will return other files called 'ptr[len]' from the path.
+ *
+ * Only on the first call 'ptr' and 'len' are used.  For repeating calls they
+ * don't need valid values.
+ *
+ * If nothing found on the first call the option FNAME_MESS will issue the
+ * message:
+ *	    'Can't find file "<file>" in path'
+ * On repeating calls:
+ *	    'No more file "<file>" found in path'
+ *
+ * options:
+ * FNAME_MESS	    give error message when not found
+ *
+ * Uses NameBuff[]!
+ *
+ * Returns an allocated string for the file name.  NULL for error.
+ *
+ */
+    char_u *
+find_file_in_path(
+    char_u	*ptr,		// file name
+    int		len,		// length of file name
+    int		options,
+    int		first,		// use count'th matching file name
+    char_u	*rel_fname)	// file name searching relative to
+{
+    return find_file_in_path_option(ptr, len, options, first,
+	    *curbuf->b_p_path == NUL ? p_path : curbuf->b_p_path,
+	    FINDFILE_BOTH, rel_fname, curbuf->b_p_sua);
+}
+
+static char_u	*ff_file_to_find = NULL;
+static void	*fdip_search_ctx = NULL;
+
+# if defined(EXITFREE) || defined(PROTO)
+    void
+free_findfile(void)
+{
+    vim_free(ff_file_to_find);
+    vim_findfile_cleanup(fdip_search_ctx);
+    vim_free(ff_expand_buffer);
+}
+# endif
+
+/*
+ * Find the directory name "ptr[len]" in the path.
+ *
+ * options:
+ * FNAME_MESS	    give error message when not found
+ * FNAME_UNESC	    unescape backslashes.
+ *
+ * Uses NameBuff[]!
+ *
+ * Returns an allocated string for the file name.  NULL for error.
+ */
+    char_u *
+find_directory_in_path(
+    char_u	*ptr,		// file name
+    int		len,		// length of file name
+    int		options,
+    char_u	*rel_fname)	// file name searching relative to
+{
+    return find_file_in_path_option(ptr, len, options, TRUE, p_cdpath,
+				       FINDFILE_DIR, rel_fname, (char_u *)"");
+}
+
+    char_u *
+find_file_in_path_option(
+    char_u	*ptr,		// file name
+    int		len,		// length of file name
+    int		options,
+    int		first,		// use count'th matching file name
+    char_u	*path_option,	// p_path or p_cdpath
+    int		find_what,	// FINDFILE_FILE, _DIR or _BOTH
+    char_u	*rel_fname,	// file name we are looking relative to.
+    char_u	*suffixes)	// list of suffixes, 'suffixesadd' option
+{
+    static char_u	*dir;
+    static int		did_findfile_init = FALSE;
+    char_u		save_char;
+    char_u		*file_name = NULL;
+    char_u		*buf = NULL;
+    int			rel_to_curdir;
+# ifdef AMIGA
+    struct Process	*proc = (struct Process *)FindTask(0L);
+    APTR		save_winptr = proc->pr_WindowPtr;
+
+    // Avoid a requester here for a volume that doesn't exist.
+    proc->pr_WindowPtr = (APTR)-1L;
+# endif
+
+    if (first == TRUE)
+    {
+	// copy file name into NameBuff, expanding environment variables
+	save_char = ptr[len];
+	ptr[len] = NUL;
+	expand_env_esc(ptr, NameBuff, MAXPATHL, FALSE, TRUE, NULL);
+	ptr[len] = save_char;
+
+	vim_free(ff_file_to_find);
+	ff_file_to_find = vim_strsave(NameBuff);
+	if (ff_file_to_find == NULL)	// out of memory
+	{
+	    file_name = NULL;
+	    goto theend;
+	}
+	if (options & FNAME_UNESC)
+	{
+	    // Change all "\ " to " ".
+	    for (ptr = ff_file_to_find; *ptr != NUL; ++ptr)
+		if (ptr[0] == '\\' && ptr[1] == ' ')
+		    mch_memmove(ptr, ptr + 1, STRLEN(ptr));
+	}
+    }
+
+    rel_to_curdir = (ff_file_to_find[0] == '.'
+		    && (ff_file_to_find[1] == NUL
+			|| vim_ispathsep(ff_file_to_find[1])
+			|| (ff_file_to_find[1] == '.'
+			    && (ff_file_to_find[2] == NUL
+				|| vim_ispathsep(ff_file_to_find[2])))));
+    if (vim_isAbsName(ff_file_to_find)
+	    // "..", "../path", "." and "./path": don't use the path_option
+	    || rel_to_curdir
+# if defined(MSWIN)
+	    // handle "\tmp" as absolute path
+	    || vim_ispathsep(ff_file_to_find[0])
+	    // handle "c:name" as absolute path
+	    || (ff_file_to_find[0] != NUL && ff_file_to_find[1] == ':')
+# endif
+# ifdef AMIGA
+	    // handle ":tmp" as absolute path
+	    || ff_file_to_find[0] == ':'
+# endif
+       )
+    {
+	/*
+	 * Absolute path, no need to use "path_option".
+	 * If this is not a first call, return NULL.  We already returned a
+	 * filename on the first call.
+	 */
+	if (first == TRUE)
+	{
+	    int		l;
+	    int		run;
+
+	    if (path_with_url(ff_file_to_find))
+	    {
+		file_name = vim_strsave(ff_file_to_find);
+		goto theend;
+	    }
+
+	    // When FNAME_REL flag given first use the directory of the file.
+	    // Otherwise or when this fails use the current directory.
+	    for (run = 1; run <= 2; ++run)
+	    {
+		l = (int)STRLEN(ff_file_to_find);
+		if (run == 1
+			&& rel_to_curdir
+			&& (options & FNAME_REL)
+			&& rel_fname != NULL
+			&& STRLEN(rel_fname) + l < MAXPATHL)
+		{
+		    STRCPY(NameBuff, rel_fname);
+		    STRCPY(gettail(NameBuff), ff_file_to_find);
+		    l = (int)STRLEN(NameBuff);
+		}
+		else
+		{
+		    STRCPY(NameBuff, ff_file_to_find);
+		    run = 2;
+		}
+
+		// When the file doesn't exist, try adding parts of
+		// 'suffixesadd'.
+		buf = suffixes;
+		for (;;)
+		{
+		    if (mch_getperm(NameBuff) >= 0
+			     && (find_what == FINDFILE_BOTH
+				 || ((find_what == FINDFILE_DIR)
+						    == mch_isdir(NameBuff))))
+		    {
+			file_name = vim_strsave(NameBuff);
+			goto theend;
+		    }
+		    if (*buf == NUL)
+			break;
+		    copy_option_part(&buf, NameBuff + l, MAXPATHL - l, ",");
+		}
+	    }
+	}
+    }
+    else
+    {
+	/*
+	 * Loop over all paths in the 'path' or 'cdpath' option.
+	 * When "first" is set, first setup to the start of the option.
+	 * Otherwise continue to find the next match.
+	 */
+	if (first == TRUE)
+	{
+	    // vim_findfile_free_visited can handle a possible NULL pointer
+	    vim_findfile_free_visited(fdip_search_ctx);
+	    dir = path_option;
+	    did_findfile_init = FALSE;
+	}
+
+	for (;;)
+	{
+	    if (did_findfile_init)
+	    {
+		file_name = vim_findfile(fdip_search_ctx);
+		if (file_name != NULL)
+		    break;
+
+		did_findfile_init = FALSE;
+	    }
+	    else
+	    {
+		char_u  *r_ptr;
+
+		if (dir == NULL || *dir == NUL)
+		{
+		    // We searched all paths of the option, now we can
+		    // free the search context.
+		    vim_findfile_cleanup(fdip_search_ctx);
+		    fdip_search_ctx = NULL;
+		    break;
+		}
+
+		if ((buf = alloc(MAXPATHL)) == NULL)
+		    break;
+
+		// copy next path
+		buf[0] = 0;
+		copy_option_part(&dir, buf, MAXPATHL, " ,");
+
+# ifdef FEAT_PATH_EXTRA
+		// get the stopdir string
+		r_ptr = vim_findfile_stopdir(buf);
+# else
+		r_ptr = NULL;
+# endif
+		fdip_search_ctx = vim_findfile_init(buf, ff_file_to_find,
+					    r_ptr, 100, FALSE, find_what,
+					   fdip_search_ctx, FALSE, rel_fname);
+		if (fdip_search_ctx != NULL)
+		    did_findfile_init = TRUE;
+		vim_free(buf);
+	    }
+	}
+    }
+    if (file_name == NULL && (options & FNAME_MESS))
+    {
+	if (first == TRUE)
+	{
+	    if (find_what == FINDFILE_DIR)
+		semsg(_("E344: Can't find directory \"%s\" in cdpath"),
+			ff_file_to_find);
+	    else
+		semsg(_("E345: Can't find file \"%s\" in path"),
+			ff_file_to_find);
+	}
+	else
+	{
+	    if (find_what == FINDFILE_DIR)
+		semsg(_("E346: No more directory \"%s\" found in cdpath"),
+			ff_file_to_find);
+	    else
+		semsg(_("E347: No more file \"%s\" found in path"),
+			ff_file_to_find);
+	}
+    }
+
+theend:
+# ifdef AMIGA
+    proc->pr_WindowPtr = save_winptr;
+# endif
+    return file_name;
+}
+
+/*
+ * Get the file name at the cursor.
+ * If Visual mode is active, use the selected text if it's in one line.
+ * Returns the name in allocated memory, NULL for failure.
+ */
+    char_u *
+grab_file_name(long count, linenr_T *file_lnum)
+{
+    int options = FNAME_MESS|FNAME_EXP|FNAME_REL|FNAME_UNESC;
+
+    if (VIsual_active)
+    {
+	int	len;
+	char_u	*ptr;
+
+	if (get_visual_text(NULL, &ptr, &len) == FAIL)
+	    return NULL;
+	// Only recognize ":123" here
+	if (file_lnum != NULL && ptr[len] == ':' && isdigit(ptr[len + 1]))
+	{
+	    char_u *p = ptr + len + 1;
+
+	    *file_lnum = getdigits(&p);
+	}
+	return find_file_name_in_path(ptr, len, options,
+						     count, curbuf->b_ffname);
+    }
+    return file_name_at_cursor(options | FNAME_HYP, count, file_lnum);
+}
+
+/*
+ * Return the file name under or after the cursor.
+ *
+ * The 'path' option is searched if the file name is not absolute.
+ * The string returned has been alloc'ed and should be freed by the caller.
+ * NULL is returned if the file name or file is not found.
+ *
+ * options:
+ * FNAME_MESS	    give error messages
+ * FNAME_EXP	    expand to path
+ * FNAME_HYP	    check for hypertext link
+ * FNAME_INCL	    apply "includeexpr"
+ */
+    char_u *
+file_name_at_cursor(int options, long count, linenr_T *file_lnum)
+{
+    return file_name_in_line(ml_get_curline(),
+		      curwin->w_cursor.col, options, count, curbuf->b_ffname,
+		      file_lnum);
+}
+
+/*
+ * Return the name of the file under or after ptr[col].
+ * Otherwise like file_name_at_cursor().
+ */
+    char_u *
+file_name_in_line(
+    char_u	*line,
+    int		col,
+    int		options,
+    long	count,
+    char_u	*rel_fname,	// file we are searching relative to
+    linenr_T	*file_lnum)	// line number after the file name
+{
+    char_u	*ptr;
+    int		len;
+    int		in_type = TRUE;
+    int		is_url = FALSE;
+
+    /*
+     * search forward for what could be the start of a file name
+     */
+    ptr = line + col;
+    while (*ptr != NUL && !vim_isfilec(*ptr))
+	MB_PTR_ADV(ptr);
+    if (*ptr == NUL)		// nothing found
+    {
+	if (options & FNAME_MESS)
+	    emsg(_("E446: No file name under cursor"));
+	return NULL;
+    }
+
+    /*
+     * Search backward for first char of the file name.
+     * Go one char back to ":" before "//" even when ':' is not in 'isfname'.
+     */
+    while (ptr > line)
+    {
+	if (has_mbyte && (len = (*mb_head_off)(line, ptr - 1)) > 0)
+	    ptr -= len + 1;
+	else if (vim_isfilec(ptr[-1])
+		|| ((options & FNAME_HYP) && path_is_url(ptr - 1)))
+	    --ptr;
+	else
+	    break;
+    }
+
+    /*
+     * Search forward for the last char of the file name.
+     * Also allow "://" when ':' is not in 'isfname'.
+     */
+    len = 0;
+    while (vim_isfilec(ptr[len]) || (ptr[len] == '\\' && ptr[len + 1] == ' ')
+			 || ((options & FNAME_HYP) && path_is_url(ptr + len))
+			 || (is_url && vim_strchr((char_u *)":?&=", ptr[len]) != NULL))
+    {
+	// After type:// we also include :, ?, & and = as valid characters, so that
+	// http://google.com:8080?q=this&that=ok works.
+	if ((ptr[len] >= 'A' && ptr[len] <= 'Z') || (ptr[len] >= 'a' && ptr[len] <= 'z'))
+	{
+	    if (in_type && path_is_url(ptr + len + 1))
+		is_url = TRUE;
+	}
+	else
+	    in_type = FALSE;
+
+	if (ptr[len] == '\\')
+	    // Skip over the "\" in "\ ".
+	    ++len;
+	if (has_mbyte)
+	    len += (*mb_ptr2len)(ptr + len);
+	else
+	    ++len;
+    }
+
+    /*
+     * If there is trailing punctuation, remove it.
+     * But don't remove "..", could be a directory name.
+     */
+    if (len > 2 && vim_strchr((char_u *)".,:;!", ptr[len - 1]) != NULL
+						       && ptr[len - 2] != '.')
+	--len;
+
+    if (file_lnum != NULL)
+    {
+	char_u *p;
+	char	*line_english = " line ";
+	char	*line_transl = _(line_msg);
+
+	// Get the number after the file name and a separator character.
+	// Also accept " line 999" with and without the same translation as
+	// used in last_set_msg().
+	p = ptr + len;
+	if (STRNCMP(p, line_english, STRLEN(line_english)) == 0)
+	    p += STRLEN(line_english);
+	else if (STRNCMP(p, line_transl, STRLEN(line_transl)) == 0)
+	    p += STRLEN(line_transl);
+	else
+	    p = skipwhite(p);
+	if (*p != NUL)
+	{
+	    if (!isdigit(*p))
+		++p;		    // skip the separator
+	    p = skipwhite(p);
+	    if (isdigit(*p))
+		*file_lnum = (int)getdigits(&p);
+	}
+    }
+
+    return find_file_name_in_path(ptr, len, options, count, rel_fname);
+}
+
+# if defined(FEAT_FIND_ID) && defined(FEAT_EVAL)
+    static char_u *
+eval_includeexpr(char_u *ptr, int len)
+{
+    char_u	*res;
+
+    set_vim_var_string(VV_FNAME, ptr, len);
+    res = eval_to_string_safe(curbuf->b_p_inex,
+		      was_set_insecurely((char_u *)"includeexpr", OPT_LOCAL));
+    set_vim_var_string(VV_FNAME, NULL, 0);
+    return res;
+}
+# endif
+
+/*
+ * Return the name of the file ptr[len] in 'path'.
+ * Otherwise like file_name_at_cursor().
+ */
+    char_u *
+find_file_name_in_path(
+    char_u	*ptr,
+    int		len,
+    int		options,
+    long	count,
+    char_u	*rel_fname)	// file we are searching relative to
+{
+    char_u	*file_name;
+    int		c;
+# if defined(FEAT_FIND_ID) && defined(FEAT_EVAL)
+    char_u	*tofree = NULL;
+
+    if ((options & FNAME_INCL) && *curbuf->b_p_inex != NUL)
+    {
+	tofree = eval_includeexpr(ptr, len);
+	if (tofree != NULL)
+	{
+	    ptr = tofree;
+	    len = (int)STRLEN(ptr);
+	}
+    }
+# endif
+
+    if (options & FNAME_EXP)
+    {
+	file_name = find_file_in_path(ptr, len, options & ~FNAME_MESS,
+							     TRUE, rel_fname);
+
+# if defined(FEAT_FIND_ID) && defined(FEAT_EVAL)
+	/*
+	 * If the file could not be found in a normal way, try applying
+	 * 'includeexpr' (unless done already).
+	 */
+	if (file_name == NULL
+		&& !(options & FNAME_INCL) && *curbuf->b_p_inex != NUL)
+	{
+	    tofree = eval_includeexpr(ptr, len);
+	    if (tofree != NULL)
+	    {
+		ptr = tofree;
+		len = (int)STRLEN(ptr);
+		file_name = find_file_in_path(ptr, len, options & ~FNAME_MESS,
+							     TRUE, rel_fname);
+	    }
+	}
+# endif
+	if (file_name == NULL && (options & FNAME_MESS))
+	{
+	    c = ptr[len];
+	    ptr[len] = NUL;
+	    semsg(_("E447: Can't find file \"%s\" in path"), ptr);
+	    ptr[len] = c;
+	}
+
+	// Repeat finding the file "count" times.  This matters when it
+	// appears several times in the path.
+	while (file_name != NULL && --count > 0)
+	{
+	    vim_free(file_name);
+	    file_name = find_file_in_path(ptr, len, options, FALSE, rel_fname);
+	}
+    }
+    else
+	file_name = vim_strnsave(ptr, len);
+
+# if defined(FEAT_FIND_ID) && defined(FEAT_EVAL)
+    vim_free(tofree);
+# endif
+
+    return file_name;
+}
+
+/*
+ * Return the end of the directory name, on the first path
+ * separator:
+ * "/path/file", "/path/dir/", "/path//dir", "/file"
+ *	 ^	       ^	     ^	      ^
+ */
+    static char_u *
+gettail_dir(char_u *fname)
+{
+    char_u	*dir_end = fname;
+    char_u	*next_dir_end = fname;
+    int		look_for_sep = TRUE;
+    char_u	*p;
+
+    for (p = fname; *p != NUL; )
+    {
+	if (vim_ispathsep(*p))
+	{
+	    if (look_for_sep)
+	    {
+		next_dir_end = p;
+		look_for_sep = FALSE;
+	    }
+	}
+	else
+	{
+	    if (!look_for_sep)
+		dir_end = next_dir_end;
+	    look_for_sep = TRUE;
+	}
+	MB_PTR_ADV(p);
+    }
+    return dir_end;
+}
+
+/*
+ * return TRUE if 'c' is a path list separator.
+ */
+    int
+vim_ispathlistsep(int c)
+{
+# ifdef UNIX
+    return (c == ':');
+# else
+    return (c == ';');	// might not be right for every system...
+# endif
+}
+
+/*
+ * Moves "*psep" back to the previous path separator in "path".
+ * Returns FAIL is "*psep" ends up at the beginning of "path".
+ */
+    static int
+find_previous_pathsep(char_u *path, char_u **psep)
+{
+    // skip the current separator
+    if (*psep > path && vim_ispathsep(**psep))
+	--*psep;
+
+    // find the previous separator
+    while (*psep > path)
+    {
+	if (vim_ispathsep(**psep))
+	    return OK;
+	MB_PTR_BACK(path, *psep);
+    }
+
+    return FAIL;
+}
+
+/*
+ * Returns TRUE if "maybe_unique" is unique wrt other_paths in "gap".
+ * "maybe_unique" is the end portion of "((char_u **)gap->ga_data)[i]".
+ */
+    static int
+is_unique(char_u *maybe_unique, garray_T *gap, int i)
+{
+    int	    j;
+    int	    candidate_len;
+    int	    other_path_len;
+    char_u  **other_paths = (char_u **)gap->ga_data;
+    char_u  *rival;
+
+    for (j = 0; j < gap->ga_len; j++)
+    {
+	if (j == i)
+	    continue;  // don't compare it with itself
+
+	candidate_len = (int)STRLEN(maybe_unique);
+	other_path_len = (int)STRLEN(other_paths[j]);
+	if (other_path_len < candidate_len)
+	    continue;  // it's different when it's shorter
+
+	rival = other_paths[j] + other_path_len - candidate_len;
+	if (fnamecmp(maybe_unique, rival) == 0
+		&& (rival == other_paths[j] || vim_ispathsep(*(rival - 1))))
+	    return FALSE;  // match
+    }
+
+    return TRUE;  // no match found
+}
+
+/*
+ * Split the 'path' option into an array of strings in garray_T.  Relative
+ * paths are expanded to their equivalent fullpath.  This includes the "."
+ * (relative to current buffer directory) and empty path (relative to current
+ * directory) notations.
+ *
+ * TODO: handle upward search (;) and path limiter (**N) notations by
+ * expanding each into their equivalent path(s).
+ */
+    static void
+expand_path_option(char_u *curdir, garray_T *gap)
+{
+    char_u	*path_option = *curbuf->b_p_path == NUL
+						  ? p_path : curbuf->b_p_path;
+    char_u	*buf;
+    char_u	*p;
+    int		len;
+
+    if ((buf = alloc(MAXPATHL)) == NULL)
+	return;
+
+    while (*path_option != NUL)
+    {
+	copy_option_part(&path_option, buf, MAXPATHL, " ,");
+
+	if (buf[0] == '.' && (buf[1] == NUL || vim_ispathsep(buf[1])))
+	{
+	    // Relative to current buffer:
+	    // "/path/file" + "." -> "/path/"
+	    // "/path/file"  + "./subdir" -> "/path/subdir"
+	    if (curbuf->b_ffname == NULL)
+		continue;
+	    p = gettail(curbuf->b_ffname);
+	    len = (int)(p - curbuf->b_ffname);
+	    if (len + (int)STRLEN(buf) >= MAXPATHL)
+		continue;
+	    if (buf[1] == NUL)
+		buf[len] = NUL;
+	    else
+		STRMOVE(buf + len, buf + 2);
+	    mch_memmove(buf, curbuf->b_ffname, len);
+	    simplify_filename(buf);
+	}
+	else if (buf[0] == NUL)
+	    // relative to current directory
+	    STRCPY(buf, curdir);
+	else if (path_with_url(buf))
+	    // URL can't be used here
+	    continue;
+	else if (!mch_isFullName(buf))
+	{
+	    // Expand relative path to their full path equivalent
+	    len = (int)STRLEN(curdir);
+	    if (len + (int)STRLEN(buf) + 3 > MAXPATHL)
+		continue;
+	    STRMOVE(buf + len + 1, buf);
+	    STRCPY(buf, curdir);
+	    buf[len] = PATHSEP;
+	    simplify_filename(buf);
+	}
+
+	if (ga_grow(gap, 1) == FAIL)
+	    break;
+
+# if defined(MSWIN)
+	// Avoid the path ending in a backslash, it fails when a comma is
+	// appended.
+	len = (int)STRLEN(buf);
+	if (buf[len - 1] == '\\')
+	    buf[len - 1] = '/';
+# endif
+
+	p = vim_strsave(buf);
+	if (p == NULL)
+	    break;
+	((char_u **)gap->ga_data)[gap->ga_len++] = p;
+    }
+
+    vim_free(buf);
+}
+
+/*
+ * Returns a pointer to the file or directory name in "fname" that matches the
+ * longest path in "ga"p, or NULL if there is no match. For example:
+ *
+ *    path: /foo/bar/baz
+ *   fname: /foo/bar/baz/quux.txt
+ * returns:		 ^this
+ */
+    static char_u *
+get_path_cutoff(char_u *fname, garray_T *gap)
+{
+    int	    i;
+    int	    maxlen = 0;
+    char_u  **path_part = (char_u **)gap->ga_data;
+    char_u  *cutoff = NULL;
+
+    for (i = 0; i < gap->ga_len; i++)
+    {
+	int j = 0;
+
+	while ((fname[j] == path_part[i][j]
+# if defined(MSWIN)
+		|| (vim_ispathsep(fname[j]) && vim_ispathsep(path_part[i][j]))
+# endif
+			     ) && fname[j] != NUL && path_part[i][j] != NUL)
+	    j++;
+	if (j > maxlen)
+	{
+	    maxlen = j;
+	    cutoff = &fname[j];
+	}
+    }
+
+    // skip to the file or directory name
+    if (cutoff != NULL)
+	while (vim_ispathsep(*cutoff))
+	    MB_PTR_ADV(cutoff);
+
+    return cutoff;
+}
+
+/*
+ * Sorts, removes duplicates and modifies all the fullpath names in "gap" so
+ * that they are unique with respect to each other while conserving the part
+ * that matches the pattern. Beware, this is at least O(n^2) wrt "gap->ga_len".
+ */
+    void
+uniquefy_paths(garray_T *gap, char_u *pattern)
+{
+    int		i;
+    int		len;
+    char_u	**fnames = (char_u **)gap->ga_data;
+    int		sort_again = FALSE;
+    char_u	*pat;
+    char_u      *file_pattern;
+    char_u	*curdir;
+    regmatch_T	regmatch;
+    garray_T	path_ga;
+    char_u	**in_curdir = NULL;
+    char_u	*short_name;
+
+    remove_duplicates(gap);
+    ga_init2(&path_ga, (int)sizeof(char_u *), 1);
+
+    /*
+     * We need to prepend a '*' at the beginning of file_pattern so that the
+     * regex matches anywhere in the path. FIXME: is this valid for all
+     * possible patterns?
+     */
+    len = (int)STRLEN(pattern);
+    file_pattern = alloc(len + 2);
+    if (file_pattern == NULL)
+	return;
+    file_pattern[0] = '*';
+    file_pattern[1] = NUL;
+    STRCAT(file_pattern, pattern);
+    pat = file_pat_to_reg_pat(file_pattern, NULL, NULL, TRUE);
+    vim_free(file_pattern);
+    if (pat == NULL)
+	return;
+
+    regmatch.rm_ic = TRUE;		// always ignore case
+    regmatch.regprog = vim_regcomp(pat, RE_MAGIC + RE_STRING);
+    vim_free(pat);
+    if (regmatch.regprog == NULL)
+	return;
+
+    if ((curdir = alloc(MAXPATHL)) == NULL)
+	goto theend;
+    mch_dirname(curdir, MAXPATHL);
+    expand_path_option(curdir, &path_ga);
+
+    in_curdir = ALLOC_CLEAR_MULT(char_u *, gap->ga_len);
+    if (in_curdir == NULL)
+	goto theend;
+
+    for (i = 0; i < gap->ga_len && !got_int; i++)
+    {
+	char_u	    *path = fnames[i];
+	int	    is_in_curdir;
+	char_u	    *dir_end = gettail_dir(path);
+	char_u	    *pathsep_p;
+	char_u	    *path_cutoff;
+
+	len = (int)STRLEN(path);
+	is_in_curdir = fnamencmp(curdir, path, dir_end - path) == 0
+					     && curdir[dir_end - path] == NUL;
+	if (is_in_curdir)
+	    in_curdir[i] = vim_strsave(path);
+
+	// Shorten the filename while maintaining its uniqueness
+	path_cutoff = get_path_cutoff(path, &path_ga);
+
+	// Don't assume all files can be reached without path when search
+	// pattern starts with star star slash, so only remove path_cutoff
+	// when possible.
+	if (pattern[0] == '*' && pattern[1] == '*'
+		&& vim_ispathsep_nocolon(pattern[2])
+		&& path_cutoff != NULL
+		&& vim_regexec(&regmatch, path_cutoff, (colnr_T)0)
+		&& is_unique(path_cutoff, gap, i))
+	{
+	    sort_again = TRUE;
+	    mch_memmove(path, path_cutoff, STRLEN(path_cutoff) + 1);
+	}
+	else
+	{
+	    // Here all files can be reached without path, so get shortest
+	    // unique path.  We start at the end of the path.
+	    pathsep_p = path + len - 1;
+
+	    while (find_previous_pathsep(path, &pathsep_p))
+		if (vim_regexec(&regmatch, pathsep_p + 1, (colnr_T)0)
+			&& is_unique(pathsep_p + 1, gap, i)
+			&& path_cutoff != NULL && pathsep_p + 1 >= path_cutoff)
+		{
+		    sort_again = TRUE;
+		    mch_memmove(path, pathsep_p + 1, STRLEN(pathsep_p));
+		    break;
+		}
+	}
+
+	if (mch_isFullName(path))
+	{
+	    /*
+	     * Last resort: shorten relative to curdir if possible.
+	     * 'possible' means:
+	     * 1. It is under the current directory.
+	     * 2. The result is actually shorter than the original.
+	     *
+	     *	    Before		  curdir	After
+	     *	    /foo/bar/file.txt	  /foo/bar	./file.txt
+	     *	    c:\foo\bar\file.txt   c:\foo\bar	.\file.txt
+	     *	    /file.txt		  /		/file.txt
+	     *	    c:\file.txt		  c:\		.\file.txt
+	     */
+	    short_name = shorten_fname(path, curdir);
+	    if (short_name != NULL && short_name > path + 1
+# if defined(MSWIN)
+		    // On windows,
+		    //	    shorten_fname("c:\a\a.txt", "c:\a\b")
+		    // returns "\a\a.txt", which is not really the short
+		    // name, hence:
+		    && !vim_ispathsep(*short_name)
+# endif
+		)
+	    {
+		STRCPY(path, ".");
+		add_pathsep(path);
+		STRMOVE(path + STRLEN(path), short_name);
+	    }
+	}
+	ui_breakcheck();
+    }
+
+    // Shorten filenames in /in/current/directory/{filename}
+    for (i = 0; i < gap->ga_len && !got_int; i++)
+    {
+	char_u *rel_path;
+	char_u *path = in_curdir[i];
+
+	if (path == NULL)
+	    continue;
+
+	// If the {filename} is not unique, change it to ./{filename}.
+	// Else reduce it to {filename}
+	short_name = shorten_fname(path, curdir);
+	if (short_name == NULL)
+	    short_name = path;
+	if (is_unique(short_name, gap, i))
+	{
+	    STRCPY(fnames[i], short_name);
+	    continue;
+	}
+
+	rel_path = alloc(STRLEN(short_name) + STRLEN(PATHSEPSTR) + 2);
+	if (rel_path == NULL)
+	    goto theend;
+	STRCPY(rel_path, ".");
+	add_pathsep(rel_path);
+	STRCAT(rel_path, short_name);
+
+	vim_free(fnames[i]);
+	fnames[i] = rel_path;
+	sort_again = TRUE;
+	ui_breakcheck();
+    }
+
+theend:
+    vim_free(curdir);
+    if (in_curdir != NULL)
+    {
+	for (i = 0; i < gap->ga_len; i++)
+	    vim_free(in_curdir[i]);
+	vim_free(in_curdir);
+    }
+    ga_clear_strings(&path_ga);
+    vim_regfree(regmatch.regprog);
+
+    if (sort_again)
+	remove_duplicates(gap);
+}
+
+/*
+ * Calls globpath() with 'path' values for the given pattern and stores the
+ * result in "gap".
+ * Returns the total number of matches.
+ */
+    int
+expand_in_path(
+    garray_T	*gap,
+    char_u	*pattern,
+    int		flags)		// EW_* flags
+{
+    char_u	*curdir;
+    garray_T	path_ga;
+    char_u	*paths = NULL;
+    int		glob_flags = 0;
+
+    if ((curdir = alloc(MAXPATHL)) == NULL)
+	return 0;
+    mch_dirname(curdir, MAXPATHL);
+
+    ga_init2(&path_ga, (int)sizeof(char_u *), 1);
+    expand_path_option(curdir, &path_ga);
+    vim_free(curdir);
+    if (path_ga.ga_len == 0)
+	return 0;
+
+    paths = ga_concat_strings(&path_ga, ",");
+    ga_clear_strings(&path_ga);
+    if (paths == NULL)
+	return 0;
+
+    if (flags & EW_ICASE)
+	glob_flags |= WILD_ICASE;
+    if (flags & EW_ADDSLASH)
+	glob_flags |= WILD_ADD_SLASH;
+    globpath(paths, pattern, gap, glob_flags);
+    vim_free(paths);
+
+    return gap->ga_len;
+}
+
+#endif // FEAT_SEARCHPATH
+
+/*
+ * Converts a file name into a canonical form. It simplifies a file name into
+ * its simplest form by stripping out unneeded components, if any.  The
+ * resulting file name is simplified in place and will either be the same
+ * length as that supplied, or shorter.
+ */
+    void
+simplify_filename(char_u *filename)
+{
+#ifndef AMIGA	    // Amiga doesn't have "..", it uses "/"
+    int		components = 0;
+    char_u	*p, *tail, *start;
+    int		stripping_disabled = FALSE;
+    int		relative = TRUE;
+
+    p = filename;
+# ifdef BACKSLASH_IN_FILENAME
+    if (p[1] == ':')	    // skip "x:"
+	p += 2;
+# endif
+
+    if (vim_ispathsep(*p))
+    {
+	relative = FALSE;
+	do
+	    ++p;
+	while (vim_ispathsep(*p));
+    }
+    start = p;	    // remember start after "c:/" or "/" or "///"
+#ifdef UNIX
+    // Posix says that "//path" is unchanged but "///path" is "/path".
+    if (start > filename + 2)
+    {
+	STRMOVE(filename + 1, p);
+	start = p = filename + 1;
+    }
+#endif
+
+    do
+    {
+	// At this point "p" is pointing to the char following a single "/"
+	// or "p" is at the "start" of the (absolute or relative) path name.
+# ifdef VMS
+	// VMS allows device:[path] - don't strip the [ in directory
+	if ((*p == '[' || *p == '<') && p > filename && p[-1] == ':')
+	{
+	    // :[ or :< composition: vms directory component
+	    ++components;
+	    p = getnextcomp(p + 1);
+	}
+	// allow remote calls as host"user passwd"::device:[path]
+	else if (p[0] == ':' && p[1] == ':' && p > filename && p[-1] == '"' )
+	{
+	    // ":: composition: vms host/passwd component
+	    ++components;
+	    p = getnextcomp(p + 2);
+	}
+	else
+# endif
+	  if (vim_ispathsep(*p))
+	    STRMOVE(p, p + 1);		// remove duplicate "/"
+	else if (p[0] == '.' && (vim_ispathsep(p[1]) || p[1] == NUL))
+	{
+	    if (p == start && relative)
+		p += 1 + (p[1] != NUL);	// keep single "." or leading "./"
+	    else
+	    {
+		// Strip "./" or ".///".  If we are at the end of the file name
+		// and there is no trailing path separator, either strip "/." if
+		// we are after "start", or strip "." if we are at the beginning
+		// of an absolute path name .
+		tail = p + 1;
+		if (p[1] != NUL)
+		    while (vim_ispathsep(*tail))
+			MB_PTR_ADV(tail);
+		else if (p > start)
+		    --p;		// strip preceding path separator
+		STRMOVE(p, tail);
+	    }
+	}
+	else if (p[0] == '.' && p[1] == '.' &&
+	    (vim_ispathsep(p[2]) || p[2] == NUL))
+	{
+	    // Skip to after ".." or "../" or "..///".
+	    tail = p + 2;
+	    while (vim_ispathsep(*tail))
+		MB_PTR_ADV(tail);
+
+	    if (components > 0)		// strip one preceding component
+	    {
+		int		do_strip = FALSE;
+		char_u		saved_char;
+		stat_T		st;
+
+		// Don't strip for an erroneous file name.
+		if (!stripping_disabled)
+		{
+		    // If the preceding component does not exist in the file
+		    // system, we strip it.  On Unix, we don't accept a symbolic
+		    // link that refers to a non-existent file.
+		    saved_char = p[-1];
+		    p[-1] = NUL;
+# ifdef UNIX
+		    if (mch_lstat((char *)filename, &st) < 0)
+# else
+			if (mch_stat((char *)filename, &st) < 0)
+# endif
+			    do_strip = TRUE;
+		    p[-1] = saved_char;
+
+		    --p;
+		    // Skip back to after previous '/'.
+		    while (p > start && !after_pathsep(start, p))
+			MB_PTR_BACK(start, p);
+
+		    if (!do_strip)
+		    {
+			// If the component exists in the file system, check
+			// that stripping it won't change the meaning of the
+			// file name.  First get information about the
+			// unstripped file name.  This may fail if the component
+			// to strip is not a searchable directory (but a regular
+			// file, for instance), since the trailing "/.." cannot
+			// be applied then.  We don't strip it then since we
+			// don't want to replace an erroneous file name by
+			// a valid one, and we disable stripping of later
+			// components.
+			saved_char = *tail;
+			*tail = NUL;
+			if (mch_stat((char *)filename, &st) >= 0)
+			    do_strip = TRUE;
+			else
+			    stripping_disabled = TRUE;
+			*tail = saved_char;
+# ifdef UNIX
+			if (do_strip)
+			{
+			    stat_T	new_st;
+
+			    // On Unix, the check for the unstripped file name
+			    // above works also for a symbolic link pointing to
+			    // a searchable directory.  But then the parent of
+			    // the directory pointed to by the link must be the
+			    // same as the stripped file name.  (The latter
+			    // exists in the file system since it is the
+			    // component's parent directory.)
+			    if (p == start && relative)
+				(void)mch_stat(".", &new_st);
+			    else
+			    {
+				saved_char = *p;
+				*p = NUL;
+				(void)mch_stat((char *)filename, &new_st);
+				*p = saved_char;
+			    }
+
+			    if (new_st.st_ino != st.st_ino ||
+				new_st.st_dev != st.st_dev)
+			    {
+				do_strip = FALSE;
+				// We don't disable stripping of later
+				// components since the unstripped path name is
+				// still valid.
+			    }
+			}
+# endif
+		    }
+		}
+
+		if (!do_strip)
+		{
+		    // Skip the ".." or "../" and reset the counter for the
+		    // components that might be stripped later on.
+		    p = tail;
+		    components = 0;
+		}
+		else
+		{
+		    // Strip previous component.  If the result would get empty
+		    // and there is no trailing path separator, leave a single
+		    // "." instead.  If we are at the end of the file name and
+		    // there is no trailing path separator and a preceding
+		    // component is left after stripping, strip its trailing
+		    // path separator as well.
+		    if (p == start && relative && tail[-1] == '.')
+		    {
+			*p++ = '.';
+			*p = NUL;
+		    }
+		    else
+		    {
+			if (p > start && tail[-1] == '.')
+			    --p;
+			STRMOVE(p, tail);	// strip previous component
+		    }
+
+		    --components;
+		}
+	    }
+	    else if (p == start && !relative)	// leading "/.." or "/../"
+		STRMOVE(p, tail);		// strip ".." or "../"
+	    else
+	    {
+		if (p == start + 2 && p[-2] == '.')	// leading "./../"
+		{
+		    STRMOVE(p - 2, p);			// strip leading "./"
+		    tail -= 2;
+		}
+		p = tail;		// skip to char after ".." or "../"
+	    }
+	}
+	else
+	{
+	    ++components;		// simple path component
+	    p = getnextcomp(p);
+	}
+    } while (*p != NUL);
+#endif // !AMIGA
+}
+
+#if defined(FEAT_EVAL) || defined(PROTO)
+/*
+ * "simplify()" function
+ */
+    void
+f_simplify(typval_T *argvars, typval_T *rettv)
+{
+    char_u	*p;
+
+    p = tv_get_string(&argvars[0]);
+    rettv->vval.v_string = vim_strsave(p);
+    simplify_filename(rettv->vval.v_string);	// simplify in place
+    rettv->v_type = VAR_STRING;
+}
+#endif // FEAT_EVAL