xref: /vim-8.2.3635/src/mbyte.c (revision 577fadfc)
1 /* vi:set ts=8 sts=4 sw=4 noet:
2  *
3  * VIM - Vi IMproved	by Bram Moolenaar
4  * Multibyte extensions partly by Sung-Hoon Baek
5  *
6  * Do ":help uganda"  in Vim to read copying and usage conditions.
7  * Do ":help credits" in Vim to see a list of people who contributed.
8  * See README.txt for an overview of the Vim source code.
9  */
10 /*
11  * mbyte.c: Code specifically for handling multi-byte characters.
12  *
13  * The encoding used in the core is set with 'encoding'.  When 'encoding' is
14  * changed, the following four variables are set (for speed).
15  * Currently these types of character encodings are supported:
16  *
17  * "enc_dbcs"	    When non-zero it tells the type of double byte character
18  *		    encoding (Chinese, Korean, Japanese, etc.).
19  *		    The cell width on the display is equal to the number of
20  *		    bytes.  (exception: DBCS_JPNU with first byte 0x8e)
21  *		    Recognizing the first or second byte is difficult, it
22  *		    requires checking a byte sequence from the start.
23  * "enc_utf8"	    When TRUE use Unicode characters in UTF-8 encoding.
24  *		    The cell width on the display needs to be determined from
25  *		    the character value.
26  *		    Recognizing bytes is easy: 0xxx.xxxx is a single-byte
27  *		    char, 10xx.xxxx is a trailing byte, 11xx.xxxx is a leading
28  *		    byte of a multi-byte character.
29  *		    To make things complicated, up to six composing characters
30  *		    are allowed.  These are drawn on top of the first char.
31  *		    For most editing the sequence of bytes with composing
32  *		    characters included is considered to be one character.
33  * "enc_unicode"    When 2 use 16-bit Unicode characters (or UTF-16).
34  *		    When 4 use 32-but Unicode characters.
35  *		    Internally characters are stored in UTF-8 encoding to
36  *		    avoid NUL bytes.  Conversion happens when doing I/O.
37  *		    "enc_utf8" will also be TRUE.
38  *
39  * "has_mbyte" is set when "enc_dbcs" or "enc_utf8" is non-zero.
40  *
41  * If none of these is TRUE, 8-bit bytes are used for a character.  The
42  * encoding isn't currently specified (TODO).
43  *
44  * 'encoding' specifies the encoding used in the core.  This is in registers,
45  * text manipulation, buffers, etc.  Conversion has to be done when characters
46  * in another encoding are received or send:
47  *
48  *		       clipboard
49  *			   ^
50  *			   | (2)
51  *			   V
52  *		   +---------------+
53  *	      (1)  |		   | (3)
54  *  keyboard ----->|	 core	   |-----> display
55  *		   |		   |
56  *		   +---------------+
57  *			   ^
58  *			   | (4)
59  *			   V
60  *			 file
61  *
62  * (1) Typed characters arrive in the current locale.  Conversion is to be
63  *     done when 'encoding' is different from 'termencoding'.
64  * (2) Text will be made available with the encoding specified with
65  *     'encoding'.  If this is not sufficient, system-specific conversion
66  *     might be required.
67  * (3) For the GUI the correct font must be selected, no conversion done.
68  *     Otherwise, conversion is to be done when 'encoding' differs from
69  *     'termencoding'.  (Different in the GTK+ 2 port -- 'termencoding'
70  *     is always used for both input and output and must always be set to
71  *     "utf-8".  gui_mch_init() does this automatically.)
72  * (4) The encoding of the file is specified with 'fileencoding'.  Conversion
73  *     is to be done when it's different from 'encoding'.
74  *
75  * The viminfo file is a special case: Only text is converted, not file names.
76  * Vim scripts may contain an ":encoding" command.  This has an effect for
77  * some commands, like ":menutrans"
78  */
79 
80 #include "vim.h"
81 
82 #ifdef WIN32UNIX
83 # ifndef WIN32_LEAN_AND_MEAN
84 #  define WIN32_LEAN_AND_MEAN
85 # endif
86 # if defined(FEAT_GUI) || defined(FEAT_XCLIPBOARD)
87 #  include <X11/Xwindows.h>
88 #  define WINBYTE wBYTE
89 # else
90 #  include <windows.h>
91 #  define WINBYTE BYTE
92 # endif
93 # ifdef WIN32
94 #  undef WIN32	    /* Some windows.h define WIN32, we don't want that here. */
95 # endif
96 #else
97 # define WINBYTE BYTE
98 #endif
99 
100 #if (defined(MSWIN) || defined(WIN32UNIX)) && !defined(__MINGW32__)
101 # include <winnls.h>
102 #endif
103 
104 #ifdef FEAT_GUI_X11
105 # include <X11/Intrinsic.h>
106 #endif
107 #ifdef X_LOCALE
108 # include <X11/Xlocale.h>
109 # if !defined(HAVE_MBLEN) && !defined(mblen)
110 #  define mblen _Xmblen
111 # endif
112 #endif
113 
114 #if defined(FEAT_GUI_GTK) && defined(FEAT_XIM)
115 # if GTK_CHECK_VERSION(3,0,0)
116 #  include <gdk/gdkkeysyms-compat.h>
117 # else
118 #  include <gdk/gdkkeysyms.h>
119 # endif
120 # ifdef MSWIN
121 #  include <gdk/gdkwin32.h>
122 # else
123 #  include <gdk/gdkx.h>
124 # endif
125 #endif
126 
127 #ifdef HAVE_WCHAR_H
128 # include <wchar.h>
129 #endif
130 
131 #if 0
132 /* This has been disabled, because several people reported problems with the
133  * wcwidth() and iswprint() library functions, esp. for Hebrew. */
134 # ifdef __STDC_ISO_10646__
135 #  define USE_WCHAR_FUNCTIONS
136 # endif
137 #endif
138 
139 static int dbcs_char2len(int c);
140 static int dbcs_char2bytes(int c, char_u *buf);
141 static int dbcs_ptr2len(char_u *p);
142 static int dbcs_ptr2len_len(char_u *p, int size);
143 static int utf_ptr2cells_len(char_u *p, int size);
144 static int dbcs_char2cells(int c);
145 static int dbcs_ptr2cells_len(char_u *p, int size);
146 static int dbcs_ptr2char(char_u *p);
147 
148 /*
149  * Lookup table to quickly get the length in bytes of a UTF-8 character from
150  * the first byte of a UTF-8 string.
151  * Bytes which are illegal when used as the first byte have a 1.
152  * The NUL byte has length 1.
153  */
154 static char utf8len_tab[256] =
155 {
156     1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,
157     1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,
158     1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,
159     1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,
160     1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,
161     1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,
162     2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,
163     3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,4,4,4,4,4,4,4,4,5,5,5,5,6,6,1,1,
164 };
165 
166 /*
167  * Like utf8len_tab above, but using a zero for illegal lead bytes.
168  */
169 static char utf8len_tab_zero[256] =
170 {
171     1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,
172     1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,
173     1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,
174     1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,
175     0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
176     0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
177     2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,
178     3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,4,4,4,4,4,4,4,4,5,5,5,5,6,6,0,0,
179 };
180 
181 /*
182  * XIM often causes trouble.  Define XIM_DEBUG to get a log of XIM callbacks
183  * in the "xim.log" file.
184  */
185 /* #define XIM_DEBUG */
186 #ifdef XIM_DEBUG
187     static void
188 xim_log(char *s, ...)
189 {
190     va_list arglist;
191     static FILE *fd = NULL;
192 
193     if (fd == (FILE *)-1)
194 	return;
195     if (fd == NULL)
196     {
197 	fd = mch_fopen("xim.log", "w");
198 	if (fd == NULL)
199 	{
200 	    emsg("Cannot open xim.log");
201 	    fd = (FILE *)-1;
202 	    return;
203 	}
204     }
205 
206     va_start(arglist, s);
207     vfprintf(fd, s, arglist);
208     va_end(arglist);
209 }
210 #endif
211 
212 
213 /*
214  * Canonical encoding names and their properties.
215  * "iso-8859-n" is handled by enc_canonize() directly.
216  */
217 static struct
218 {   char *name;		int prop;		int codepage;}
219 enc_canon_table[] =
220 {
221 #define IDX_LATIN_1	0
222     {"latin1",		ENC_8BIT + ENC_LATIN1,	1252},
223 #define IDX_ISO_2	1
224     {"iso-8859-2",	ENC_8BIT,		0},
225 #define IDX_ISO_3	2
226     {"iso-8859-3",	ENC_8BIT,		0},
227 #define IDX_ISO_4	3
228     {"iso-8859-4",	ENC_8BIT,		0},
229 #define IDX_ISO_5	4
230     {"iso-8859-5",	ENC_8BIT,		0},
231 #define IDX_ISO_6	5
232     {"iso-8859-6",	ENC_8BIT,		0},
233 #define IDX_ISO_7	6
234     {"iso-8859-7",	ENC_8BIT,		0},
235 #define IDX_ISO_8	7
236     {"iso-8859-8",	ENC_8BIT,		0},
237 #define IDX_ISO_9	8
238     {"iso-8859-9",	ENC_8BIT,		0},
239 #define IDX_ISO_10	9
240     {"iso-8859-10",	ENC_8BIT,		0},
241 #define IDX_ISO_11	10
242     {"iso-8859-11",	ENC_8BIT,		0},
243 #define IDX_ISO_13	11
244     {"iso-8859-13",	ENC_8BIT,		0},
245 #define IDX_ISO_14	12
246     {"iso-8859-14",	ENC_8BIT,		0},
247 #define IDX_ISO_15	13
248     {"iso-8859-15",	ENC_8BIT + ENC_LATIN9,	0},
249 #define IDX_KOI8_R	14
250     {"koi8-r",		ENC_8BIT,		0},
251 #define IDX_KOI8_U	15
252     {"koi8-u",		ENC_8BIT,		0},
253 #define IDX_UTF8	16
254     {"utf-8",		ENC_UNICODE,		0},
255 #define IDX_UCS2	17
256     {"ucs-2",		ENC_UNICODE + ENC_ENDIAN_B + ENC_2BYTE, 0},
257 #define IDX_UCS2LE	18
258     {"ucs-2le",		ENC_UNICODE + ENC_ENDIAN_L + ENC_2BYTE, 0},
259 #define IDX_UTF16	19
260     {"utf-16",		ENC_UNICODE + ENC_ENDIAN_B + ENC_2WORD, 0},
261 #define IDX_UTF16LE	20
262     {"utf-16le",	ENC_UNICODE + ENC_ENDIAN_L + ENC_2WORD, 0},
263 #define IDX_UCS4	21
264     {"ucs-4",		ENC_UNICODE + ENC_ENDIAN_B + ENC_4BYTE, 0},
265 #define IDX_UCS4LE	22
266     {"ucs-4le",		ENC_UNICODE + ENC_ENDIAN_L + ENC_4BYTE, 0},
267 
268     /* For debugging DBCS encoding on Unix. */
269 #define IDX_DEBUG	23
270     {"debug",		ENC_DBCS,		DBCS_DEBUG},
271 #define IDX_EUC_JP	24
272     {"euc-jp",		ENC_DBCS,		DBCS_JPNU},
273 #define IDX_SJIS	25
274     {"sjis",		ENC_DBCS,		DBCS_JPN},
275 #define IDX_EUC_KR	26
276     {"euc-kr",		ENC_DBCS,		DBCS_KORU},
277 #define IDX_EUC_CN	27
278     {"euc-cn",		ENC_DBCS,		DBCS_CHSU},
279 #define IDX_EUC_TW	28
280     {"euc-tw",		ENC_DBCS,		DBCS_CHTU},
281 #define IDX_BIG5	29
282     {"big5",		ENC_DBCS,		DBCS_CHT},
283 
284     /* MS-DOS and MS-Windows codepages are included here, so that they can be
285      * used on Unix too.  Most of them are similar to ISO-8859 encodings, but
286      * not exactly the same. */
287 #define IDX_CP437	30
288     {"cp437",		ENC_8BIT,		437}, /* like iso-8859-1 */
289 #define IDX_CP737	31
290     {"cp737",		ENC_8BIT,		737}, /* like iso-8859-7 */
291 #define IDX_CP775	32
292     {"cp775",		ENC_8BIT,		775}, /* Baltic */
293 #define IDX_CP850	33
294     {"cp850",		ENC_8BIT,		850}, /* like iso-8859-4 */
295 #define IDX_CP852	34
296     {"cp852",		ENC_8BIT,		852}, /* like iso-8859-1 */
297 #define IDX_CP855	35
298     {"cp855",		ENC_8BIT,		855}, /* like iso-8859-2 */
299 #define IDX_CP857	36
300     {"cp857",		ENC_8BIT,		857}, /* like iso-8859-5 */
301 #define IDX_CP860	37
302     {"cp860",		ENC_8BIT,		860}, /* like iso-8859-9 */
303 #define IDX_CP861	38
304     {"cp861",		ENC_8BIT,		861}, /* like iso-8859-1 */
305 #define IDX_CP862	39
306     {"cp862",		ENC_8BIT,		862}, /* like iso-8859-1 */
307 #define IDX_CP863	40
308     {"cp863",		ENC_8BIT,		863}, /* like iso-8859-8 */
309 #define IDX_CP865	41
310     {"cp865",		ENC_8BIT,		865}, /* like iso-8859-1 */
311 #define IDX_CP866	42
312     {"cp866",		ENC_8BIT,		866}, /* like iso-8859-5 */
313 #define IDX_CP869	43
314     {"cp869",		ENC_8BIT,		869}, /* like iso-8859-7 */
315 #define IDX_CP874	44
316     {"cp874",		ENC_8BIT,		874}, /* Thai */
317 #define IDX_CP932	45
318     {"cp932",		ENC_DBCS,		DBCS_JPN},
319 #define IDX_CP936	46
320     {"cp936",		ENC_DBCS,		DBCS_CHS},
321 #define IDX_CP949	47
322     {"cp949",		ENC_DBCS,		DBCS_KOR},
323 #define IDX_CP950	48
324     {"cp950",		ENC_DBCS,		DBCS_CHT},
325 #define IDX_CP1250	49
326     {"cp1250",		ENC_8BIT,		1250}, /* Czech, Polish, etc. */
327 #define IDX_CP1251	50
328     {"cp1251",		ENC_8BIT,		1251}, /* Cyrillic */
329     /* cp1252 is considered to be equal to latin1 */
330 #define IDX_CP1253	51
331     {"cp1253",		ENC_8BIT,		1253}, /* Greek */
332 #define IDX_CP1254	52
333     {"cp1254",		ENC_8BIT,		1254}, /* Turkish */
334 #define IDX_CP1255	53
335     {"cp1255",		ENC_8BIT,		1255}, /* Hebrew */
336 #define IDX_CP1256	54
337     {"cp1256",		ENC_8BIT,		1256}, /* Arabic */
338 #define IDX_CP1257	55
339     {"cp1257",		ENC_8BIT,		1257}, /* Baltic */
340 #define IDX_CP1258	56
341     {"cp1258",		ENC_8BIT,		1258}, /* Vietnamese */
342 
343 #define IDX_MACROMAN	57
344     {"macroman",	ENC_8BIT + ENC_MACROMAN, 0},	/* Mac OS */
345 #define IDX_DECMCS	58
346     {"dec-mcs",		ENC_8BIT,		0},	/* DEC MCS */
347 #define IDX_HPROMAN8	59
348     {"hp-roman8",	ENC_8BIT,		0},	/* HP Roman8 */
349 #define IDX_COUNT	60
350 };
351 
352 /*
353  * Aliases for encoding names.
354  */
355 static struct
356 {   char *name;		int canon;}
357 enc_alias_table[] =
358 {
359     {"ansi",		IDX_LATIN_1},
360     {"iso-8859-1",	IDX_LATIN_1},
361     {"latin2",		IDX_ISO_2},
362     {"latin3",		IDX_ISO_3},
363     {"latin4",		IDX_ISO_4},
364     {"cyrillic",	IDX_ISO_5},
365     {"arabic",		IDX_ISO_6},
366     {"greek",		IDX_ISO_7},
367 #ifdef MSWIN
368     {"hebrew",		IDX_CP1255},
369 #else
370     {"hebrew",		IDX_ISO_8},
371 #endif
372     {"latin5",		IDX_ISO_9},
373     {"turkish",		IDX_ISO_9}, /* ? */
374     {"latin6",		IDX_ISO_10},
375     {"nordic",		IDX_ISO_10}, /* ? */
376     {"thai",		IDX_ISO_11}, /* ? */
377     {"latin7",		IDX_ISO_13},
378     {"latin8",		IDX_ISO_14},
379     {"latin9",		IDX_ISO_15},
380     {"utf8",		IDX_UTF8},
381     {"unicode",		IDX_UCS2},
382     {"ucs2",		IDX_UCS2},
383     {"ucs2be",		IDX_UCS2},
384     {"ucs-2be",		IDX_UCS2},
385     {"ucs2le",		IDX_UCS2LE},
386     {"utf16",		IDX_UTF16},
387     {"utf16be",		IDX_UTF16},
388     {"utf-16be",	IDX_UTF16},
389     {"utf16le",		IDX_UTF16LE},
390     {"ucs4",		IDX_UCS4},
391     {"ucs4be",		IDX_UCS4},
392     {"ucs-4be",		IDX_UCS4},
393     {"ucs4le",		IDX_UCS4LE},
394     {"utf32",		IDX_UCS4},
395     {"utf-32",		IDX_UCS4},
396     {"utf32be",		IDX_UCS4},
397     {"utf-32be",	IDX_UCS4},
398     {"utf32le",		IDX_UCS4LE},
399     {"utf-32le",	IDX_UCS4LE},
400     {"932",		IDX_CP932},
401     {"949",		IDX_CP949},
402     {"936",		IDX_CP936},
403     {"gbk",		IDX_CP936},
404     {"950",		IDX_CP950},
405     {"eucjp",		IDX_EUC_JP},
406     {"unix-jis",	IDX_EUC_JP},
407     {"ujis",		IDX_EUC_JP},
408     {"shift-jis",	IDX_SJIS},
409     {"pck",		IDX_SJIS},	/* Sun: PCK */
410     {"euckr",		IDX_EUC_KR},
411     {"5601",		IDX_EUC_KR},	/* Sun: KS C 5601 */
412     {"euccn",		IDX_EUC_CN},
413     {"gb2312",		IDX_EUC_CN},
414     {"euctw",		IDX_EUC_TW},
415 #if defined(MSWIN) || defined(WIN32UNIX) || defined(MACOS_X)
416     {"japan",		IDX_CP932},
417     {"korea",		IDX_CP949},
418     {"prc",		IDX_CP936},
419     {"chinese",		IDX_CP936},
420     {"taiwan",		IDX_CP950},
421     {"big5",		IDX_CP950},
422 #else
423     {"japan",		IDX_EUC_JP},
424     {"korea",		IDX_EUC_KR},
425     {"prc",		IDX_EUC_CN},
426     {"chinese",		IDX_EUC_CN},
427     {"taiwan",		IDX_EUC_TW},
428     {"cp950",		IDX_BIG5},
429     {"950",		IDX_BIG5},
430 #endif
431     {"mac",		IDX_MACROMAN},
432     {"mac-roman",	IDX_MACROMAN},
433     {NULL,		0}
434 };
435 
436 #ifndef CP_UTF8
437 # define CP_UTF8 65001	/* magic number from winnls.h */
438 #endif
439 
440 /*
441  * Find encoding "name" in the list of canonical encoding names.
442  * Returns -1 if not found.
443  */
444     static int
445 enc_canon_search(char_u *name)
446 {
447     int		i;
448 
449     for (i = 0; i < IDX_COUNT; ++i)
450 	if (STRCMP(name, enc_canon_table[i].name) == 0)
451 	    return i;
452     return -1;
453 }
454 
455 
456 /*
457  * Find canonical encoding "name" in the list and return its properties.
458  * Returns 0 if not found.
459  */
460     int
461 enc_canon_props(char_u *name)
462 {
463     int		i;
464 
465     i = enc_canon_search(name);
466     if (i >= 0)
467 	return enc_canon_table[i].prop;
468 #ifdef MSWIN
469     if (name[0] == 'c' && name[1] == 'p' && VIM_ISDIGIT(name[2]))
470     {
471 	CPINFO	cpinfo;
472 
473 	/* Get info on this codepage to find out what it is. */
474 	if (GetCPInfo(atoi((char *)name + 2), &cpinfo) != 0)
475 	{
476 	    if (cpinfo.MaxCharSize == 1) /* some single-byte encoding */
477 		return ENC_8BIT;
478 	    if (cpinfo.MaxCharSize == 2
479 		    && (cpinfo.LeadByte[0] != 0 || cpinfo.LeadByte[1] != 0))
480 		/* must be a DBCS encoding */
481 		return ENC_DBCS;
482 	}
483 	return 0;
484     }
485 #endif
486     if (STRNCMP(name, "2byte-", 6) == 0)
487 	return ENC_DBCS;
488     if (STRNCMP(name, "8bit-", 5) == 0 || STRNCMP(name, "iso-8859-", 9) == 0)
489 	return ENC_8BIT;
490     return 0;
491 }
492 
493 /*
494  * Set up for using multi-byte characters.
495  * Called in three cases:
496  * - by main() to initialize (p_enc == NULL)
497  * - by set_init_1() after 'encoding' was set to its default.
498  * - by do_set() when 'encoding' has been set.
499  * p_enc must have been passed through enc_canonize() already.
500  * Sets the "enc_unicode", "enc_utf8", "enc_dbcs" and "has_mbyte" flags.
501  * Fills mb_bytelen_tab[] and returns NULL when there are no problems.
502  * When there is something wrong: Returns an error message and doesn't change
503  * anything.
504  */
505     char *
506 mb_init(void)
507 {
508     int		i;
509     int		idx;
510     int		n;
511     int		enc_dbcs_new = 0;
512 #if defined(USE_ICONV) && !defined(MSWIN) && !defined(WIN32UNIX) \
513 	&& !defined(MACOS_CONVERT)
514 # define LEN_FROM_CONV
515     vimconv_T	vimconv;
516     char_u	*p;
517 #endif
518 
519     if (p_enc == NULL)
520     {
521 	/* Just starting up: set the whole table to one's. */
522 	for (i = 0; i < 256; ++i)
523 	    mb_bytelen_tab[i] = 1;
524 	input_conv.vc_type = CONV_NONE;
525 	input_conv.vc_factor = 1;
526 	output_conv.vc_type = CONV_NONE;
527 	return NULL;
528     }
529 
530 #ifdef MSWIN
531     if (p_enc[0] == 'c' && p_enc[1] == 'p' && VIM_ISDIGIT(p_enc[2]))
532     {
533 	CPINFO	cpinfo;
534 
535 	/* Get info on this codepage to find out what it is. */
536 	if (GetCPInfo(atoi((char *)p_enc + 2), &cpinfo) != 0)
537 	{
538 	    if (cpinfo.MaxCharSize == 1)
539 	    {
540 		/* some single-byte encoding */
541 		enc_unicode = 0;
542 		enc_utf8 = FALSE;
543 	    }
544 	    else if (cpinfo.MaxCharSize == 2
545 		    && (cpinfo.LeadByte[0] != 0 || cpinfo.LeadByte[1] != 0))
546 	    {
547 		/* must be a DBCS encoding, check below */
548 		enc_dbcs_new = atoi((char *)p_enc + 2);
549 	    }
550 	    else
551 		goto codepage_invalid;
552 	}
553 	else if (GetLastError() == ERROR_INVALID_PARAMETER)
554 	{
555 codepage_invalid:
556 	    return N_("E543: Not a valid codepage");
557 	}
558     }
559 #endif
560     else if (STRNCMP(p_enc, "8bit-", 5) == 0
561 	    || STRNCMP(p_enc, "iso-8859-", 9) == 0)
562     {
563 	/* Accept any "8bit-" or "iso-8859-" name. */
564 	enc_unicode = 0;
565 	enc_utf8 = FALSE;
566     }
567     else if (STRNCMP(p_enc, "2byte-", 6) == 0)
568     {
569 #ifdef MSWIN
570 	/* Windows: accept only valid codepage numbers, check below. */
571 	if (p_enc[6] != 'c' || p_enc[7] != 'p'
572 			      || (enc_dbcs_new = atoi((char *)p_enc + 8)) == 0)
573 	    return e_invarg;
574 #else
575 	/* Unix: accept any "2byte-" name, assume current locale. */
576 	enc_dbcs_new = DBCS_2BYTE;
577 #endif
578     }
579     else if ((idx = enc_canon_search(p_enc)) >= 0)
580     {
581 	i = enc_canon_table[idx].prop;
582 	if (i & ENC_UNICODE)
583 	{
584 	    /* Unicode */
585 	    enc_utf8 = TRUE;
586 	    if (i & (ENC_2BYTE | ENC_2WORD))
587 		enc_unicode = 2;
588 	    else if (i & ENC_4BYTE)
589 		enc_unicode = 4;
590 	    else
591 		enc_unicode = 0;
592 	}
593 	else if (i & ENC_DBCS)
594 	{
595 	    /* 2byte, handle below */
596 	    enc_dbcs_new = enc_canon_table[idx].codepage;
597 	}
598 	else
599 	{
600 	    /* Must be 8-bit. */
601 	    enc_unicode = 0;
602 	    enc_utf8 = FALSE;
603 	}
604     }
605     else    /* Don't know what encoding this is, reject it. */
606 	return e_invarg;
607 
608     if (enc_dbcs_new != 0)
609     {
610 #ifdef MSWIN
611 	/* Check if the DBCS code page is OK. */
612 	if (!IsValidCodePage(enc_dbcs_new))
613 	    goto codepage_invalid;
614 #endif
615 	enc_unicode = 0;
616 	enc_utf8 = FALSE;
617     }
618     enc_dbcs = enc_dbcs_new;
619     has_mbyte = (enc_dbcs != 0 || enc_utf8);
620 
621 #if defined(MSWIN) || defined(FEAT_CYGWIN_WIN32_CLIPBOARD)
622     enc_codepage = encname2codepage(p_enc);
623     enc_latin9 = (STRCMP(p_enc, "iso-8859-15") == 0);
624 #endif
625 
626     /* Detect an encoding that uses latin1 characters. */
627     enc_latin1like = (enc_utf8 || STRCMP(p_enc, "latin1") == 0
628 					|| STRCMP(p_enc, "iso-8859-15") == 0);
629 
630     /*
631      * Set the function pointers.
632      */
633     if (enc_utf8)
634     {
635 	mb_ptr2len = utfc_ptr2len;
636 	mb_ptr2len_len = utfc_ptr2len_len;
637 	mb_char2len = utf_char2len;
638 	mb_char2bytes = utf_char2bytes;
639 	mb_ptr2cells = utf_ptr2cells;
640 	mb_ptr2cells_len = utf_ptr2cells_len;
641 	mb_char2cells = utf_char2cells;
642 	mb_off2cells = utf_off2cells;
643 	mb_ptr2char = utf_ptr2char;
644 	mb_head_off = utf_head_off;
645     }
646     else if (enc_dbcs != 0)
647     {
648 	mb_ptr2len = dbcs_ptr2len;
649 	mb_ptr2len_len = dbcs_ptr2len_len;
650 	mb_char2len = dbcs_char2len;
651 	mb_char2bytes = dbcs_char2bytes;
652 	mb_ptr2cells = dbcs_ptr2cells;
653 	mb_ptr2cells_len = dbcs_ptr2cells_len;
654 	mb_char2cells = dbcs_char2cells;
655 	mb_off2cells = dbcs_off2cells;
656 	mb_ptr2char = dbcs_ptr2char;
657 	mb_head_off = dbcs_head_off;
658     }
659     else
660     {
661 	mb_ptr2len = latin_ptr2len;
662 	mb_ptr2len_len = latin_ptr2len_len;
663 	mb_char2len = latin_char2len;
664 	mb_char2bytes = latin_char2bytes;
665 	mb_ptr2cells = latin_ptr2cells;
666 	mb_ptr2cells_len = latin_ptr2cells_len;
667 	mb_char2cells = latin_char2cells;
668 	mb_off2cells = latin_off2cells;
669 	mb_ptr2char = latin_ptr2char;
670 	mb_head_off = latin_head_off;
671     }
672 
673     /*
674      * Fill the mb_bytelen_tab[] for MB_BYTE2LEN().
675      */
676 #ifdef LEN_FROM_CONV
677     /* When 'encoding' is different from the current locale mblen() won't
678      * work.  Use conversion to "utf-8" instead. */
679     vimconv.vc_type = CONV_NONE;
680     if (enc_dbcs)
681     {
682 	p = enc_locale();
683 	if (p == NULL || STRCMP(p, p_enc) != 0)
684 	{
685 	    convert_setup(&vimconv, p_enc, (char_u *)"utf-8");
686 	    vimconv.vc_fail = TRUE;
687 	}
688 	vim_free(p);
689     }
690 #endif
691 
692     for (i = 0; i < 256; ++i)
693     {
694 	/* Our own function to reliably check the length of UTF-8 characters,
695 	 * independent of mblen(). */
696 	if (enc_utf8)
697 	    n = utf8len_tab[i];
698 	else if (enc_dbcs == 0)
699 	    n = 1;
700 	else
701 	{
702 #if defined(MSWIN) || defined(WIN32UNIX)
703 	    /* enc_dbcs is set by setting 'fileencoding'.  It becomes a Windows
704 	     * CodePage identifier, which we can pass directly in to Windows
705 	     * API */
706 	    n = IsDBCSLeadByteEx(enc_dbcs, (WINBYTE)i) ? 2 : 1;
707 #else
708 # if defined(__amigaos4__) || defined(__ANDROID__) || \
709 				   !(defined(HAVE_MBLEN) || defined(X_LOCALE))
710 	    /*
711 	     * if mblen() is not available, character which MSB is turned on
712 	     * are treated as leading byte character. (note : This assumption
713 	     * is not always true.)
714 	     */
715 	    n = (i & 0x80) ? 2 : 1;
716 # else
717 	    char buf[MB_MAXBYTES + 1];
718 
719 	    if (i == NUL)	/* just in case mblen() can't handle "" */
720 		n = 1;
721 	    else
722 	    {
723 		buf[0] = i;
724 		buf[1] = 0;
725 #  ifdef LEN_FROM_CONV
726 		if (vimconv.vc_type != CONV_NONE)
727 		{
728 		    /*
729 		     * string_convert() should fail when converting the first
730 		     * byte of a double-byte character.
731 		     */
732 		    p = string_convert(&vimconv, (char_u *)buf, NULL);
733 		    if (p != NULL)
734 		    {
735 			vim_free(p);
736 			n = 1;
737 		    }
738 		    else
739 			n = 2;
740 		}
741 		else
742 #  endif
743 		{
744 		    /*
745 		     * mblen() should return -1 for invalid (means the leading
746 		     * multibyte) character.  However there are some platforms
747 		     * where mblen() returns 0 for invalid character.
748 		     * Therefore, following condition includes 0.
749 		     */
750 		    vim_ignored = mblen(NULL, 0);  // First reset the state.
751 		    if (mblen(buf, (size_t)1) <= 0)
752 			n = 2;
753 		    else
754 			n = 1;
755 		}
756 	    }
757 # endif
758 #endif
759 	}
760 
761 	mb_bytelen_tab[i] = n;
762     }
763 
764 #ifdef LEN_FROM_CONV
765     convert_setup(&vimconv, NULL, NULL);
766 #endif
767 
768     /* The cell width depends on the type of multi-byte characters. */
769     (void)init_chartab();
770 
771     /* When enc_utf8 is set or reset, (de)allocate ScreenLinesUC[] */
772     screenalloc(FALSE);
773 
774     /* When using Unicode, set default for 'fileencodings'. */
775     if (enc_utf8 && !option_was_set((char_u *)"fencs"))
776 	set_string_option_direct((char_u *)"fencs", -1,
777 		       (char_u *)"ucs-bom,utf-8,default,latin1", OPT_FREE, 0);
778 
779 #if defined(HAVE_BIND_TEXTDOMAIN_CODESET) && defined(FEAT_GETTEXT)
780     /* GNU gettext 0.10.37 supports this feature: set the codeset used for
781      * translated messages independently from the current locale. */
782     (void)bind_textdomain_codeset(VIMPACKAGE,
783 					  enc_utf8 ? "utf-8" : (char *)p_enc);
784 #endif
785 
786 #ifdef MSWIN
787     /* When changing 'encoding' while starting up, then convert the command
788      * line arguments from the active codepage to 'encoding'. */
789     if (starting != 0)
790 	fix_arg_enc();
791 #endif
792 
793     /* Fire an autocommand to let people do custom font setup. This must be
794      * after Vim has been setup for the new encoding. */
795     apply_autocmds(EVENT_ENCODINGCHANGED, NULL, (char_u *)"", FALSE, curbuf);
796 
797 #ifdef FEAT_SPELL
798     /* Need to reload spell dictionaries */
799     spell_reload();
800 #endif
801 
802     return NULL;
803 }
804 
805 /*
806  * Return the size of the BOM for the current buffer:
807  * 0 - no BOM
808  * 2 - UCS-2 or UTF-16 BOM
809  * 4 - UCS-4 BOM
810  * 3 - UTF-8 BOM
811  */
812     int
813 bomb_size(void)
814 {
815     int n = 0;
816 
817     if (curbuf->b_p_bomb && !curbuf->b_p_bin)
818     {
819 	if (*curbuf->b_p_fenc == NUL)
820 	{
821 	    if (enc_utf8)
822 	    {
823 		if (enc_unicode != 0)
824 		    n = enc_unicode;
825 		else
826 		    n = 3;
827 	    }
828 	}
829 	else if (STRCMP(curbuf->b_p_fenc, "utf-8") == 0)
830 	    n = 3;
831 	else if (STRNCMP(curbuf->b_p_fenc, "ucs-2", 5) == 0
832 		|| STRNCMP(curbuf->b_p_fenc, "utf-16", 6) == 0)
833 	    n = 2;
834 	else if (STRNCMP(curbuf->b_p_fenc, "ucs-4", 5) == 0)
835 	    n = 4;
836     }
837     return n;
838 }
839 
840 #if defined(FEAT_QUICKFIX) || defined(PROTO)
841 /*
842  * Remove all BOM from "s" by moving remaining text.
843  */
844     void
845 remove_bom(char_u *s)
846 {
847     if (enc_utf8)
848     {
849 	char_u *p = s;
850 
851 	while ((p = vim_strbyte(p, 0xef)) != NULL)
852 	{
853 	    if (p[1] == 0xbb && p[2] == 0xbf)
854 		STRMOVE(p, p + 3);
855 	    else
856 		++p;
857 	}
858     }
859 }
860 #endif
861 
862 /*
863  * Get class of pointer:
864  * 0 for blank or NUL
865  * 1 for punctuation
866  * 2 for an (ASCII) word character
867  * >2 for other word characters
868  */
869     int
870 mb_get_class(char_u *p)
871 {
872     return mb_get_class_buf(p, curbuf);
873 }
874 
875     int
876 mb_get_class_buf(char_u *p, buf_T *buf)
877 {
878     if (MB_BYTE2LEN(p[0]) == 1)
879     {
880 	if (p[0] == NUL || VIM_ISWHITE(p[0]))
881 	    return 0;
882 	if (vim_iswordc_buf(p[0], buf))
883 	    return 2;
884 	return 1;
885     }
886     if (enc_dbcs != 0 && p[0] != NUL && p[1] != NUL)
887 	return dbcs_class(p[0], p[1]);
888     if (enc_utf8)
889 	return utf_class_buf(utf_ptr2char(p), buf);
890     return 0;
891 }
892 
893 /*
894  * Get class of a double-byte character.  This always returns 3 or bigger.
895  * TODO: Should return 1 for punctuation.
896  */
897     int
898 dbcs_class(unsigned lead, unsigned trail)
899 {
900     switch (enc_dbcs)
901     {
902 	/* please add classify routine for your language in here */
903 
904 	case DBCS_JPNU:	/* ? */
905 	case DBCS_JPN:
906 	    {
907 		/* JIS code classification */
908 		unsigned char lb = lead;
909 		unsigned char tb = trail;
910 
911 		/* convert process code to JIS */
912 # if defined(MSWIN) || defined(WIN32UNIX) || defined(MACOS_X)
913 		/* process code is SJIS */
914 		if (lb <= 0x9f)
915 		    lb = (lb - 0x81) * 2 + 0x21;
916 		else
917 		    lb = (lb - 0xc1) * 2 + 0x21;
918 		if (tb <= 0x7e)
919 		    tb -= 0x1f;
920 		else if (tb <= 0x9e)
921 		    tb -= 0x20;
922 		else
923 		{
924 		    tb -= 0x7e;
925 		    lb += 1;
926 		}
927 # else
928 		/*
929 		 * XXX: Code page identification can not use with all
930 		 *	    system! So, some other encoding information
931 		 *	    will be needed.
932 		 *	    In japanese: SJIS,EUC,UNICODE,(JIS)
933 		 *	    Note that JIS-code system don't use as
934 		 *	    process code in most system because it uses
935 		 *	    escape sequences(JIS is context depend encoding).
936 		 */
937 		/* assume process code is JAPANESE-EUC */
938 		lb &= 0x7f;
939 		tb &= 0x7f;
940 # endif
941 		/* exceptions */
942 		switch (lb << 8 | tb)
943 		{
944 		    case 0x2121: /* ZENKAKU space */
945 			return 0;
946 		    case 0x2122: /* TOU-TEN (Japanese comma) */
947 		    case 0x2123: /* KU-TEN (Japanese period) */
948 		    case 0x2124: /* ZENKAKU comma */
949 		    case 0x2125: /* ZENKAKU period */
950 			return 1;
951 		    case 0x213c: /* prolongedsound handled as KATAKANA */
952 			return 13;
953 		}
954 		/* sieved by KU code */
955 		switch (lb)
956 		{
957 		    case 0x21:
958 		    case 0x22:
959 			/* special symbols */
960 			return 10;
961 		    case 0x23:
962 			/* alpha-numeric */
963 			return 11;
964 		    case 0x24:
965 			/* hiragana */
966 			return 12;
967 		    case 0x25:
968 			/* katakana */
969 			return 13;
970 		    case 0x26:
971 			/* greek */
972 			return 14;
973 		    case 0x27:
974 			/* russian */
975 			return 15;
976 		    case 0x28:
977 			/* lines */
978 			return 16;
979 		    default:
980 			/* kanji */
981 			return 17;
982 		}
983 	    }
984 
985 	case DBCS_KORU:	/* ? */
986 	case DBCS_KOR:
987 	    {
988 		/* KS code classification */
989 		unsigned char c1 = lead;
990 		unsigned char c2 = trail;
991 
992 		/*
993 		 * 20 : Hangul
994 		 * 21 : Hanja
995 		 * 22 : Symbols
996 		 * 23 : Alpha-numeric/Roman Letter (Full width)
997 		 * 24 : Hangul Letter(Alphabet)
998 		 * 25 : Roman Numeral/Greek Letter
999 		 * 26 : Box Drawings
1000 		 * 27 : Unit Symbols
1001 		 * 28 : Circled/Parenthesized Letter
1002 		 * 29 : Hiragana/Katakana
1003 		 * 30 : Cyrillic Letter
1004 		 */
1005 
1006 		if (c1 >= 0xB0 && c1 <= 0xC8)
1007 		    /* Hangul */
1008 		    return 20;
1009 #if defined(MSWIN) || defined(WIN32UNIX)
1010 		else if (c1 <= 0xA0 || c2 <= 0xA0)
1011 		    /* Extended Hangul Region : MS UHC(Unified Hangul Code) */
1012 		    /* c1: 0x81-0xA0 with c2: 0x41-0x5A, 0x61-0x7A, 0x81-0xFE
1013 		     * c1: 0xA1-0xC6 with c2: 0x41-0x5A, 0x61-0x7A, 0x81-0xA0
1014 		     */
1015 		    return 20;
1016 #endif
1017 
1018 		else if (c1 >= 0xCA && c1 <= 0xFD)
1019 		    /* Hanja */
1020 		    return 21;
1021 		else switch (c1)
1022 		{
1023 		    case 0xA1:
1024 		    case 0xA2:
1025 			/* Symbols */
1026 			return 22;
1027 		    case 0xA3:
1028 			/* Alpha-numeric */
1029 			return 23;
1030 		    case 0xA4:
1031 			/* Hangul Letter(Alphabet) */
1032 			return 24;
1033 		    case 0xA5:
1034 			/* Roman Numeral/Greek Letter */
1035 			return 25;
1036 		    case 0xA6:
1037 			/* Box Drawings */
1038 			return 26;
1039 		    case 0xA7:
1040 			/* Unit Symbols */
1041 			return 27;
1042 		    case 0xA8:
1043 		    case 0xA9:
1044 			if (c2 <= 0xAF)
1045 			    return 25;  /* Roman Letter */
1046 			else if (c2 >= 0xF6)
1047 			    return 22;  /* Symbols */
1048 			else
1049 			    /* Circled/Parenthesized Letter */
1050 			    return 28;
1051 		    case 0xAA:
1052 		    case 0xAB:
1053 			/* Hiragana/Katakana */
1054 			return 29;
1055 		    case 0xAC:
1056 			/* Cyrillic Letter */
1057 			return 30;
1058 		}
1059 	    }
1060 	default:
1061 	    break;
1062     }
1063     return 3;
1064 }
1065 
1066 /*
1067  * mb_char2len() function pointer.
1068  * Return length in bytes of character "c".
1069  * Returns 1 for a single-byte character.
1070  */
1071     int
1072 latin_char2len(int c UNUSED)
1073 {
1074     return 1;
1075 }
1076 
1077     static int
1078 dbcs_char2len(
1079     int		c)
1080 {
1081     if (c >= 0x100)
1082 	return 2;
1083     return 1;
1084 }
1085 
1086 /*
1087  * mb_char2bytes() function pointer.
1088  * Convert a character to its bytes.
1089  * Returns the length in bytes.
1090  */
1091     int
1092 latin_char2bytes(int c, char_u *buf)
1093 {
1094     buf[0] = c;
1095     return 1;
1096 }
1097 
1098     static int
1099 dbcs_char2bytes(int c, char_u *buf)
1100 {
1101     if (c >= 0x100)
1102     {
1103 	buf[0] = (unsigned)c >> 8;
1104 	buf[1] = c;
1105 	/* Never use a NUL byte, it causes lots of trouble.  It's an invalid
1106 	 * character anyway. */
1107 	if (buf[1] == NUL)
1108 	    buf[1] = '\n';
1109 	return 2;
1110     }
1111     buf[0] = c;
1112     return 1;
1113 }
1114 
1115 /*
1116  * mb_ptr2len() function pointer.
1117  * Get byte length of character at "*p" but stop at a NUL.
1118  * For UTF-8 this includes following composing characters.
1119  * Returns 0 when *p is NUL.
1120  */
1121     int
1122 latin_ptr2len(char_u *p)
1123 {
1124  return MB_BYTE2LEN(*p);
1125 }
1126 
1127     static int
1128 dbcs_ptr2len(
1129     char_u	*p)
1130 {
1131     int		len;
1132 
1133     /* Check if second byte is not missing. */
1134     len = MB_BYTE2LEN(*p);
1135     if (len == 2 && p[1] == NUL)
1136 	len = 1;
1137     return len;
1138 }
1139 
1140 /*
1141  * mb_ptr2len_len() function pointer.
1142  * Like mb_ptr2len(), but limit to read "size" bytes.
1143  * Returns 0 for an empty string.
1144  * Returns 1 for an illegal char or an incomplete byte sequence.
1145  */
1146     int
1147 latin_ptr2len_len(char_u *p, int size)
1148 {
1149     if (size < 1 || *p == NUL)
1150 	return 0;
1151     return 1;
1152 }
1153 
1154     static int
1155 dbcs_ptr2len_len(char_u *p, int size)
1156 {
1157     int		len;
1158 
1159     if (size < 1 || *p == NUL)
1160 	return 0;
1161     if (size == 1)
1162 	return 1;
1163     /* Check that second byte is not missing. */
1164     len = MB_BYTE2LEN(*p);
1165     if (len == 2 && p[1] == NUL)
1166 	len = 1;
1167     return len;
1168 }
1169 
1170 struct interval
1171 {
1172     long first;
1173     long last;
1174 };
1175 
1176 /*
1177  * Return TRUE if "c" is in "table[size / sizeof(struct interval)]".
1178  */
1179     static int
1180 intable(struct interval *table, size_t size, int c)
1181 {
1182     int mid, bot, top;
1183 
1184     /* first quick check for Latin1 etc. characters */
1185     if (c < table[0].first)
1186 	return FALSE;
1187 
1188     /* binary search in table */
1189     bot = 0;
1190     top = (int)(size / sizeof(struct interval) - 1);
1191     while (top >= bot)
1192     {
1193 	mid = (bot + top) / 2;
1194 	if (table[mid].last < c)
1195 	    bot = mid + 1;
1196 	else if (table[mid].first > c)
1197 	    top = mid - 1;
1198 	else
1199 	    return TRUE;
1200     }
1201     return FALSE;
1202 }
1203 
1204 /* Sorted list of non-overlapping intervals of East Asian Ambiguous
1205  * characters, generated with ../runtime/tools/unicode.vim. */
1206 static struct interval ambiguous[] =
1207 {
1208     {0x00a1, 0x00a1},
1209     {0x00a4, 0x00a4},
1210     {0x00a7, 0x00a8},
1211     {0x00aa, 0x00aa},
1212     {0x00ad, 0x00ae},
1213     {0x00b0, 0x00b4},
1214     {0x00b6, 0x00ba},
1215     {0x00bc, 0x00bf},
1216     {0x00c6, 0x00c6},
1217     {0x00d0, 0x00d0},
1218     {0x00d7, 0x00d8},
1219     {0x00de, 0x00e1},
1220     {0x00e6, 0x00e6},
1221     {0x00e8, 0x00ea},
1222     {0x00ec, 0x00ed},
1223     {0x00f0, 0x00f0},
1224     {0x00f2, 0x00f3},
1225     {0x00f7, 0x00fa},
1226     {0x00fc, 0x00fc},
1227     {0x00fe, 0x00fe},
1228     {0x0101, 0x0101},
1229     {0x0111, 0x0111},
1230     {0x0113, 0x0113},
1231     {0x011b, 0x011b},
1232     {0x0126, 0x0127},
1233     {0x012b, 0x012b},
1234     {0x0131, 0x0133},
1235     {0x0138, 0x0138},
1236     {0x013f, 0x0142},
1237     {0x0144, 0x0144},
1238     {0x0148, 0x014b},
1239     {0x014d, 0x014d},
1240     {0x0152, 0x0153},
1241     {0x0166, 0x0167},
1242     {0x016b, 0x016b},
1243     {0x01ce, 0x01ce},
1244     {0x01d0, 0x01d0},
1245     {0x01d2, 0x01d2},
1246     {0x01d4, 0x01d4},
1247     {0x01d6, 0x01d6},
1248     {0x01d8, 0x01d8},
1249     {0x01da, 0x01da},
1250     {0x01dc, 0x01dc},
1251     {0x0251, 0x0251},
1252     {0x0261, 0x0261},
1253     {0x02c4, 0x02c4},
1254     {0x02c7, 0x02c7},
1255     {0x02c9, 0x02cb},
1256     {0x02cd, 0x02cd},
1257     {0x02d0, 0x02d0},
1258     {0x02d8, 0x02db},
1259     {0x02dd, 0x02dd},
1260     {0x02df, 0x02df},
1261     {0x0300, 0x036f},
1262     {0x0391, 0x03a1},
1263     {0x03a3, 0x03a9},
1264     {0x03b1, 0x03c1},
1265     {0x03c3, 0x03c9},
1266     {0x0401, 0x0401},
1267     {0x0410, 0x044f},
1268     {0x0451, 0x0451},
1269     {0x2010, 0x2010},
1270     {0x2013, 0x2016},
1271     {0x2018, 0x2019},
1272     {0x201c, 0x201d},
1273     {0x2020, 0x2022},
1274     {0x2024, 0x2027},
1275     {0x2030, 0x2030},
1276     {0x2032, 0x2033},
1277     {0x2035, 0x2035},
1278     {0x203b, 0x203b},
1279     {0x203e, 0x203e},
1280     {0x2074, 0x2074},
1281     {0x207f, 0x207f},
1282     {0x2081, 0x2084},
1283     {0x20ac, 0x20ac},
1284     {0x2103, 0x2103},
1285     {0x2105, 0x2105},
1286     {0x2109, 0x2109},
1287     {0x2113, 0x2113},
1288     {0x2116, 0x2116},
1289     {0x2121, 0x2122},
1290     {0x2126, 0x2126},
1291     {0x212b, 0x212b},
1292     {0x2153, 0x2154},
1293     {0x215b, 0x215e},
1294     {0x2160, 0x216b},
1295     {0x2170, 0x2179},
1296     {0x2189, 0x2189},
1297     {0x2190, 0x2199},
1298     {0x21b8, 0x21b9},
1299     {0x21d2, 0x21d2},
1300     {0x21d4, 0x21d4},
1301     {0x21e7, 0x21e7},
1302     {0x2200, 0x2200},
1303     {0x2202, 0x2203},
1304     {0x2207, 0x2208},
1305     {0x220b, 0x220b},
1306     {0x220f, 0x220f},
1307     {0x2211, 0x2211},
1308     {0x2215, 0x2215},
1309     {0x221a, 0x221a},
1310     {0x221d, 0x2220},
1311     {0x2223, 0x2223},
1312     {0x2225, 0x2225},
1313     {0x2227, 0x222c},
1314     {0x222e, 0x222e},
1315     {0x2234, 0x2237},
1316     {0x223c, 0x223d},
1317     {0x2248, 0x2248},
1318     {0x224c, 0x224c},
1319     {0x2252, 0x2252},
1320     {0x2260, 0x2261},
1321     {0x2264, 0x2267},
1322     {0x226a, 0x226b},
1323     {0x226e, 0x226f},
1324     {0x2282, 0x2283},
1325     {0x2286, 0x2287},
1326     {0x2295, 0x2295},
1327     {0x2299, 0x2299},
1328     {0x22a5, 0x22a5},
1329     {0x22bf, 0x22bf},
1330     {0x2312, 0x2312},
1331     {0x2460, 0x24e9},
1332     {0x24eb, 0x254b},
1333     {0x2550, 0x2573},
1334     {0x2580, 0x258f},
1335     {0x2592, 0x2595},
1336     {0x25a0, 0x25a1},
1337     {0x25a3, 0x25a9},
1338     {0x25b2, 0x25b3},
1339     {0x25b6, 0x25b7},
1340     {0x25bc, 0x25bd},
1341     {0x25c0, 0x25c1},
1342     {0x25c6, 0x25c8},
1343     {0x25cb, 0x25cb},
1344     {0x25ce, 0x25d1},
1345     {0x25e2, 0x25e5},
1346     {0x25ef, 0x25ef},
1347     {0x2605, 0x2606},
1348     {0x2609, 0x2609},
1349     {0x260e, 0x260f},
1350     {0x261c, 0x261c},
1351     {0x261e, 0x261e},
1352     {0x2640, 0x2640},
1353     {0x2642, 0x2642},
1354     {0x2660, 0x2661},
1355     {0x2663, 0x2665},
1356     {0x2667, 0x266a},
1357     {0x266c, 0x266d},
1358     {0x266f, 0x266f},
1359     {0x269e, 0x269f},
1360     {0x26bf, 0x26bf},
1361     {0x26c6, 0x26cd},
1362     {0x26cf, 0x26d3},
1363     {0x26d5, 0x26e1},
1364     {0x26e3, 0x26e3},
1365     {0x26e8, 0x26e9},
1366     {0x26eb, 0x26f1},
1367     {0x26f4, 0x26f4},
1368     {0x26f6, 0x26f9},
1369     {0x26fb, 0x26fc},
1370     {0x26fe, 0x26ff},
1371     {0x273d, 0x273d},
1372     {0x2776, 0x277f},
1373     {0x2b56, 0x2b59},
1374     {0x3248, 0x324f},
1375     {0xe000, 0xf8ff},
1376     {0xfe00, 0xfe0f},
1377     {0xfffd, 0xfffd},
1378     {0x1f100, 0x1f10a},
1379     {0x1f110, 0x1f12d},
1380     {0x1f130, 0x1f169},
1381     {0x1f170, 0x1f18d},
1382     {0x1f18f, 0x1f190},
1383     {0x1f19b, 0x1f1ac},
1384     {0xe0100, 0xe01ef},
1385     {0xf0000, 0xffffd},
1386     {0x100000, 0x10fffd}
1387 };
1388 
1389 #if defined(FEAT_TERMINAL) || defined(PROTO)
1390 /*
1391  * utf_char2cells() with different argument type for libvterm.
1392  */
1393     int
1394 utf_uint2cells(UINT32_T c)
1395 {
1396     if (c >= 0x100 && utf_iscomposing((int)c))
1397 	return 0;
1398     return utf_char2cells((int)c);
1399 }
1400 #endif
1401 
1402 /*
1403  * For UTF-8 character "c" return 2 for a double-width character, 1 for others.
1404  * Returns 4 or 6 for an unprintable character.
1405  * Is only correct for characters >= 0x80.
1406  * When p_ambw is "double", return 2 for a character with East Asian Width
1407  * class 'A'(mbiguous).
1408  */
1409     int
1410 utf_char2cells(int c)
1411 {
1412     /* Sorted list of non-overlapping intervals of East Asian double width
1413      * characters, generated with ../runtime/tools/unicode.vim. */
1414     static struct interval doublewidth[] =
1415     {
1416 	{0x1100, 0x115f},
1417 	{0x231a, 0x231b},
1418 	{0x2329, 0x232a},
1419 	{0x23e9, 0x23ec},
1420 	{0x23f0, 0x23f0},
1421 	{0x23f3, 0x23f3},
1422 	{0x25fd, 0x25fe},
1423 	{0x2614, 0x2615},
1424 	{0x2648, 0x2653},
1425 	{0x267f, 0x267f},
1426 	{0x2693, 0x2693},
1427 	{0x26a1, 0x26a1},
1428 	{0x26aa, 0x26ab},
1429 	{0x26bd, 0x26be},
1430 	{0x26c4, 0x26c5},
1431 	{0x26ce, 0x26ce},
1432 	{0x26d4, 0x26d4},
1433 	{0x26ea, 0x26ea},
1434 	{0x26f2, 0x26f3},
1435 	{0x26f5, 0x26f5},
1436 	{0x26fa, 0x26fa},
1437 	{0x26fd, 0x26fd},
1438 	{0x2705, 0x2705},
1439 	{0x270a, 0x270b},
1440 	{0x2728, 0x2728},
1441 	{0x274c, 0x274c},
1442 	{0x274e, 0x274e},
1443 	{0x2753, 0x2755},
1444 	{0x2757, 0x2757},
1445 	{0x2795, 0x2797},
1446 	{0x27b0, 0x27b0},
1447 	{0x27bf, 0x27bf},
1448 	{0x2b1b, 0x2b1c},
1449 	{0x2b50, 0x2b50},
1450 	{0x2b55, 0x2b55},
1451 	{0x2e80, 0x2e99},
1452 	{0x2e9b, 0x2ef3},
1453 	{0x2f00, 0x2fd5},
1454 	{0x2ff0, 0x2ffb},
1455 	{0x3000, 0x303e},
1456 	{0x3041, 0x3096},
1457 	{0x3099, 0x30ff},
1458 	{0x3105, 0x312f},
1459 	{0x3131, 0x318e},
1460 	{0x3190, 0x31ba},
1461 	{0x31c0, 0x31e3},
1462 	{0x31f0, 0x321e},
1463 	{0x3220, 0x3247},
1464 	{0x3250, 0x32fe},
1465 	{0x3300, 0x4dbf},
1466 	{0x4e00, 0xa48c},
1467 	{0xa490, 0xa4c6},
1468 	{0xa960, 0xa97c},
1469 	{0xac00, 0xd7a3},
1470 	{0xf900, 0xfaff},
1471 	{0xfe10, 0xfe19},
1472 	{0xfe30, 0xfe52},
1473 	{0xfe54, 0xfe66},
1474 	{0xfe68, 0xfe6b},
1475 	{0xff01, 0xff60},
1476 	{0xffe0, 0xffe6},
1477 	{0x16fe0, 0x16fe1},
1478 	{0x17000, 0x187f1},
1479 	{0x18800, 0x18af2},
1480 	{0x1b000, 0x1b11e},
1481 	{0x1b170, 0x1b2fb},
1482 	{0x1f004, 0x1f004},
1483 	{0x1f0cf, 0x1f0cf},
1484 	{0x1f18e, 0x1f18e},
1485 	{0x1f191, 0x1f19a},
1486 	{0x1f200, 0x1f202},
1487 	{0x1f210, 0x1f23b},
1488 	{0x1f240, 0x1f248},
1489 	{0x1f250, 0x1f251},
1490 	{0x1f260, 0x1f265},
1491 	{0x1f300, 0x1f320},
1492 	{0x1f32d, 0x1f335},
1493 	{0x1f337, 0x1f37c},
1494 	{0x1f37e, 0x1f393},
1495 	{0x1f3a0, 0x1f3ca},
1496 	{0x1f3cf, 0x1f3d3},
1497 	{0x1f3e0, 0x1f3f0},
1498 	{0x1f3f4, 0x1f3f4},
1499 	{0x1f3f8, 0x1f43e},
1500 	{0x1f440, 0x1f440},
1501 	{0x1f442, 0x1f4fc},
1502 	{0x1f4ff, 0x1f53d},
1503 	{0x1f54b, 0x1f54e},
1504 	{0x1f550, 0x1f567},
1505 	{0x1f57a, 0x1f57a},
1506 	{0x1f595, 0x1f596},
1507 	{0x1f5a4, 0x1f5a4},
1508 	{0x1f5fb, 0x1f64f},
1509 	{0x1f680, 0x1f6c5},
1510 	{0x1f6cc, 0x1f6cc},
1511 	{0x1f6d0, 0x1f6d2},
1512 	{0x1f6eb, 0x1f6ec},
1513 	{0x1f6f4, 0x1f6f9},
1514 	{0x1f910, 0x1f93e},
1515 	{0x1f940, 0x1f970},
1516 	{0x1f973, 0x1f976},
1517 	{0x1f97a, 0x1f97a},
1518 	{0x1f97c, 0x1f9a2},
1519 	{0x1f9b0, 0x1f9b9},
1520 	{0x1f9c0, 0x1f9c2},
1521 	{0x1f9d0, 0x1f9ff},
1522 	{0x20000, 0x2fffd},
1523 	{0x30000, 0x3fffd}
1524     };
1525 
1526     /* Sorted list of non-overlapping intervals of Emoji characters that don't
1527      * have ambiguous or double width,
1528      * based on http://unicode.org/emoji/charts/emoji-list.html */
1529     static struct interval emoji_width[] =
1530     {
1531 	{0x1f1e6, 0x1f1ff},
1532 	{0x1f321, 0x1f321},
1533 	{0x1f324, 0x1f32c},
1534 	{0x1f336, 0x1f336},
1535 	{0x1f37d, 0x1f37d},
1536 	{0x1f396, 0x1f397},
1537 	{0x1f399, 0x1f39b},
1538 	{0x1f39e, 0x1f39f},
1539 	{0x1f3cb, 0x1f3ce},
1540 	{0x1f3d4, 0x1f3df},
1541 	{0x1f3f3, 0x1f3f5},
1542 	{0x1f3f7, 0x1f3f7},
1543 	{0x1f43f, 0x1f43f},
1544 	{0x1f441, 0x1f441},
1545 	{0x1f4fd, 0x1f4fd},
1546 	{0x1f549, 0x1f54a},
1547 	{0x1f56f, 0x1f570},
1548 	{0x1f573, 0x1f579},
1549 	{0x1f587, 0x1f587},
1550 	{0x1f58a, 0x1f58d},
1551 	{0x1f590, 0x1f590},
1552 	{0x1f5a5, 0x1f5a5},
1553 	{0x1f5a8, 0x1f5a8},
1554 	{0x1f5b1, 0x1f5b2},
1555 	{0x1f5bc, 0x1f5bc},
1556 	{0x1f5c2, 0x1f5c4},
1557 	{0x1f5d1, 0x1f5d3},
1558 	{0x1f5dc, 0x1f5de},
1559 	{0x1f5e1, 0x1f5e1},
1560 	{0x1f5e3, 0x1f5e3},
1561 	{0x1f5e8, 0x1f5e8},
1562 	{0x1f5ef, 0x1f5ef},
1563 	{0x1f5f3, 0x1f5f3},
1564 	{0x1f5fa, 0x1f5fa},
1565 	{0x1f6cb, 0x1f6cf},
1566 	{0x1f6e0, 0x1f6e5},
1567 	{0x1f6e9, 0x1f6e9},
1568 	{0x1f6f0, 0x1f6f0},
1569 	{0x1f6f3, 0x1f6f3}
1570     };
1571 
1572     if (c >= 0x100)
1573     {
1574 #ifdef USE_WCHAR_FUNCTIONS
1575 	/*
1576 	 * Assume the library function wcwidth() works better than our own
1577 	 * stuff.  It should return 1 for ambiguous width chars!
1578 	 */
1579 	int	n = wcwidth(c);
1580 
1581 	if (n < 0)
1582 	    return 6;		/* unprintable, displays <xxxx> */
1583 	if (n > 1)
1584 	    return n;
1585 #else
1586 	if (!utf_printable(c))
1587 	    return 6;		/* unprintable, displays <xxxx> */
1588 	if (intable(doublewidth, sizeof(doublewidth), c))
1589 	    return 2;
1590 #endif
1591 	if (p_emoji && intable(emoji_width, sizeof(emoji_width), c))
1592 	    return 2;
1593     }
1594 
1595     /* Characters below 0x100 are influenced by 'isprint' option */
1596     else if (c >= 0x80 && !vim_isprintc(c))
1597 	return 4;		/* unprintable, displays <xx> */
1598 
1599     if (c >= 0x80 && *p_ambw == 'd' && intable(ambiguous, sizeof(ambiguous), c))
1600 	return 2;
1601 
1602     return 1;
1603 }
1604 
1605 /*
1606  * mb_ptr2cells() function pointer.
1607  * Return the number of display cells character at "*p" occupies.
1608  * This doesn't take care of unprintable characters, use ptr2cells() for that.
1609  */
1610     int
1611 latin_ptr2cells(char_u *p UNUSED)
1612 {
1613     return 1;
1614 }
1615 
1616     int
1617 utf_ptr2cells(
1618     char_u	*p)
1619 {
1620     int		c;
1621 
1622     /* Need to convert to a wide character. */
1623     if (*p >= 0x80)
1624     {
1625 	c = utf_ptr2char(p);
1626 	/* An illegal byte is displayed as <xx>. */
1627 	if (utf_ptr2len(p) == 1 || c == NUL)
1628 	    return 4;
1629 	/* If the char is ASCII it must be an overlong sequence. */
1630 	if (c < 0x80)
1631 	    return char2cells(c);
1632 	return utf_char2cells(c);
1633     }
1634     return 1;
1635 }
1636 
1637     int
1638 dbcs_ptr2cells(char_u *p)
1639 {
1640     /* Number of cells is equal to number of bytes, except for euc-jp when
1641      * the first byte is 0x8e. */
1642     if (enc_dbcs == DBCS_JPNU && *p == 0x8e)
1643 	return 1;
1644     return MB_BYTE2LEN(*p);
1645 }
1646 
1647 /*
1648  * mb_ptr2cells_len() function pointer.
1649  * Like mb_ptr2cells(), but limit string length to "size".
1650  * For an empty string or truncated character returns 1.
1651  */
1652     int
1653 latin_ptr2cells_len(char_u *p UNUSED, int size UNUSED)
1654 {
1655     return 1;
1656 }
1657 
1658     static int
1659 utf_ptr2cells_len(char_u *p, int size)
1660 {
1661     int		c;
1662 
1663     /* Need to convert to a wide character. */
1664     if (size > 0 && *p >= 0x80)
1665     {
1666 	if (utf_ptr2len_len(p, size) < utf8len_tab[*p])
1667 	    return 1;  /* truncated */
1668 	c = utf_ptr2char(p);
1669 	/* An illegal byte is displayed as <xx>. */
1670 	if (utf_ptr2len(p) == 1 || c == NUL)
1671 	    return 4;
1672 	/* If the char is ASCII it must be an overlong sequence. */
1673 	if (c < 0x80)
1674 	    return char2cells(c);
1675 	return utf_char2cells(c);
1676     }
1677     return 1;
1678 }
1679 
1680     static int
1681 dbcs_ptr2cells_len(char_u *p, int size)
1682 {
1683     /* Number of cells is equal to number of bytes, except for euc-jp when
1684      * the first byte is 0x8e. */
1685     if (size <= 1 || (enc_dbcs == DBCS_JPNU && *p == 0x8e))
1686 	return 1;
1687     return MB_BYTE2LEN(*p);
1688 }
1689 
1690 /*
1691  * mb_char2cells() function pointer.
1692  * Return the number of display cells character "c" occupies.
1693  * Only takes care of multi-byte chars, not "^C" and such.
1694  */
1695     int
1696 latin_char2cells(int c UNUSED)
1697 {
1698     return 1;
1699 }
1700 
1701     static int
1702 dbcs_char2cells(int c)
1703 {
1704     /* Number of cells is equal to number of bytes, except for euc-jp when
1705      * the first byte is 0x8e. */
1706     if (enc_dbcs == DBCS_JPNU && ((unsigned)c >> 8) == 0x8e)
1707 	return 1;
1708     /* use the first byte */
1709     return MB_BYTE2LEN((unsigned)c >> 8);
1710 }
1711 
1712 /*
1713  * Return the number of cells occupied by string "p".
1714  * Stop at a NUL character.  When "len" >= 0 stop at character "p[len]".
1715  */
1716     int
1717 mb_string2cells(char_u *p, int len)
1718 {
1719     int i;
1720     int clen = 0;
1721 
1722     for (i = 0; (len < 0 || i < len) && p[i] != NUL; i += (*mb_ptr2len)(p + i))
1723 	clen += (*mb_ptr2cells)(p + i);
1724     return clen;
1725 }
1726 
1727 /*
1728  * mb_off2cells() function pointer.
1729  * Return number of display cells for char at ScreenLines[off].
1730  * We make sure that the offset used is less than "max_off".
1731  */
1732     int
1733 latin_off2cells(unsigned off UNUSED, unsigned max_off UNUSED)
1734 {
1735     return 1;
1736 }
1737 
1738     int
1739 dbcs_off2cells(unsigned off, unsigned max_off)
1740 {
1741     /* never check beyond end of the line */
1742     if (off >= max_off)
1743 	return 1;
1744 
1745     /* Number of cells is equal to number of bytes, except for euc-jp when
1746      * the first byte is 0x8e. */
1747     if (enc_dbcs == DBCS_JPNU && ScreenLines[off] == 0x8e)
1748 	return 1;
1749     return MB_BYTE2LEN(ScreenLines[off]);
1750 }
1751 
1752     int
1753 utf_off2cells(unsigned off, unsigned max_off)
1754 {
1755     return (off + 1 < max_off && ScreenLines[off + 1] == 0) ? 2 : 1;
1756 }
1757 
1758 /*
1759  * mb_ptr2char() function pointer.
1760  * Convert a byte sequence into a character.
1761  */
1762     int
1763 latin_ptr2char(char_u *p)
1764 {
1765     return *p;
1766 }
1767 
1768     static int
1769 dbcs_ptr2char(char_u *p)
1770 {
1771     if (MB_BYTE2LEN(*p) > 1 && p[1] != NUL)
1772 	return (p[0] << 8) + p[1];
1773     return *p;
1774 }
1775 
1776 /*
1777  * Convert a UTF-8 byte sequence to a wide character.
1778  * If the sequence is illegal or truncated by a NUL the first byte is
1779  * returned.
1780  * For an overlong sequence this may return zero.
1781  * Does not include composing characters, of course.
1782  */
1783     int
1784 utf_ptr2char(char_u *p)
1785 {
1786     int		len;
1787 
1788     if (p[0] < 0x80)	/* be quick for ASCII */
1789 	return p[0];
1790 
1791     len = utf8len_tab_zero[p[0]];
1792     if (len > 1 && (p[1] & 0xc0) == 0x80)
1793     {
1794 	if (len == 2)
1795 	    return ((p[0] & 0x1f) << 6) + (p[1] & 0x3f);
1796 	if ((p[2] & 0xc0) == 0x80)
1797 	{
1798 	    if (len == 3)
1799 		return ((p[0] & 0x0f) << 12) + ((p[1] & 0x3f) << 6)
1800 		    + (p[2] & 0x3f);
1801 	    if ((p[3] & 0xc0) == 0x80)
1802 	    {
1803 		if (len == 4)
1804 		    return ((p[0] & 0x07) << 18) + ((p[1] & 0x3f) << 12)
1805 			+ ((p[2] & 0x3f) << 6) + (p[3] & 0x3f);
1806 		if ((p[4] & 0xc0) == 0x80)
1807 		{
1808 		    if (len == 5)
1809 			return ((p[0] & 0x03) << 24) + ((p[1] & 0x3f) << 18)
1810 			    + ((p[2] & 0x3f) << 12) + ((p[3] & 0x3f) << 6)
1811 			    + (p[4] & 0x3f);
1812 		    if ((p[5] & 0xc0) == 0x80 && len == 6)
1813 			return ((p[0] & 0x01) << 30) + ((p[1] & 0x3f) << 24)
1814 			    + ((p[2] & 0x3f) << 18) + ((p[3] & 0x3f) << 12)
1815 			    + ((p[4] & 0x3f) << 6) + (p[5] & 0x3f);
1816 		}
1817 	    }
1818 	}
1819     }
1820     /* Illegal value, just return the first byte */
1821     return p[0];
1822 }
1823 
1824 /*
1825  * Convert a UTF-8 byte sequence to a wide character.
1826  * String is assumed to be terminated by NUL or after "n" bytes, whichever
1827  * comes first.
1828  * The function is safe in the sense that it never accesses memory beyond the
1829  * first "n" bytes of "s".
1830  *
1831  * On success, returns decoded codepoint, advances "s" to the beginning of
1832  * next character and decreases "n" accordingly.
1833  *
1834  * If end of string was reached, returns 0 and, if "n" > 0, advances "s" past
1835  * NUL byte.
1836  *
1837  * If byte sequence is illegal or incomplete, returns -1 and does not advance
1838  * "s".
1839  */
1840     static int
1841 utf_safe_read_char_adv(char_u **s, size_t *n)
1842 {
1843     int		c, k;
1844 
1845     if (*n == 0) /* end of buffer */
1846 	return 0;
1847 
1848     k = utf8len_tab_zero[**s];
1849 
1850     if (k == 1)
1851     {
1852 	/* ASCII character or NUL */
1853 	(*n)--;
1854 	return *(*s)++;
1855     }
1856 
1857     if ((size_t)k <= *n)
1858     {
1859 	/* We have a multibyte sequence and it isn't truncated by buffer
1860 	 * limits so utf_ptr2char() is safe to use. Or the first byte is
1861 	 * illegal (k=0), and it's also safe to use utf_ptr2char(). */
1862 	c = utf_ptr2char(*s);
1863 
1864 	/* On failure, utf_ptr2char() returns the first byte, so here we
1865 	 * check equality with the first byte. The only non-ASCII character
1866 	 * which equals the first byte of its own UTF-8 representation is
1867 	 * U+00C3 (UTF-8: 0xC3 0x83), so need to check that special case too.
1868 	 * It's safe even if n=1, else we would have k=2 > n. */
1869 	if (c != (int)(**s) || (c == 0xC3 && (*s)[1] == 0x83))
1870 	{
1871 	    /* byte sequence was successfully decoded */
1872 	    *s += k;
1873 	    *n -= k;
1874 	    return c;
1875 	}
1876     }
1877 
1878     /* byte sequence is incomplete or illegal */
1879     return -1;
1880 }
1881 
1882 /*
1883  * Get character at **pp and advance *pp to the next character.
1884  * Note: composing characters are skipped!
1885  */
1886     int
1887 mb_ptr2char_adv(char_u **pp)
1888 {
1889     int		c;
1890 
1891     c = (*mb_ptr2char)(*pp);
1892     *pp += (*mb_ptr2len)(*pp);
1893     return c;
1894 }
1895 
1896 /*
1897  * Get character at **pp and advance *pp to the next character.
1898  * Note: composing characters are returned as separate characters.
1899  */
1900     int
1901 mb_cptr2char_adv(char_u **pp)
1902 {
1903     int		c;
1904 
1905     c = (*mb_ptr2char)(*pp);
1906     if (enc_utf8)
1907 	*pp += utf_ptr2len(*pp);
1908     else
1909 	*pp += (*mb_ptr2len)(*pp);
1910     return c;
1911 }
1912 
1913 #if defined(FEAT_ARABIC) || defined(PROTO)
1914 /*
1915  * Check if the character pointed to by "p2" is a composing character when it
1916  * comes after "p1".  For Arabic sometimes "ab" is replaced with "c", which
1917  * behaves like a composing character.
1918  */
1919     int
1920 utf_composinglike(char_u *p1, char_u *p2)
1921 {
1922     int		c2;
1923 
1924     c2 = utf_ptr2char(p2);
1925     if (utf_iscomposing(c2))
1926 	return TRUE;
1927     if (!arabic_maycombine(c2))
1928 	return FALSE;
1929     return arabic_combine(utf_ptr2char(p1), c2);
1930 }
1931 #endif
1932 
1933 /*
1934  * Convert a UTF-8 byte string to a wide character.  Also get up to MAX_MCO
1935  * composing characters.
1936  */
1937     int
1938 utfc_ptr2char(
1939     char_u	*p,
1940     int		*pcc)	/* return: composing chars, last one is 0 */
1941 {
1942     int		len;
1943     int		c;
1944     int		cc;
1945     int		i = 0;
1946 
1947     c = utf_ptr2char(p);
1948     len = utf_ptr2len(p);
1949 
1950     /* Only accept a composing char when the first char isn't illegal. */
1951     if ((len > 1 || *p < 0x80)
1952 	    && p[len] >= 0x80
1953 	    && UTF_COMPOSINGLIKE(p, p + len))
1954     {
1955 	cc = utf_ptr2char(p + len);
1956 	for (;;)
1957 	{
1958 	    pcc[i++] = cc;
1959 	    if (i == MAX_MCO)
1960 		break;
1961 	    len += utf_ptr2len(p + len);
1962 	    if (p[len] < 0x80 || !utf_iscomposing(cc = utf_ptr2char(p + len)))
1963 		break;
1964 	}
1965     }
1966 
1967     if (i < MAX_MCO)	/* last composing char must be 0 */
1968 	pcc[i] = 0;
1969 
1970     return c;
1971 }
1972 
1973 /*
1974  * Convert a UTF-8 byte string to a wide character.  Also get up to MAX_MCO
1975  * composing characters.  Use no more than p[maxlen].
1976  */
1977     int
1978 utfc_ptr2char_len(
1979     char_u	*p,
1980     int		*pcc,	/* return: composing chars, last one is 0 */
1981     int		maxlen)
1982 {
1983     int		len;
1984     int		c;
1985     int		cc;
1986     int		i = 0;
1987 
1988     c = utf_ptr2char(p);
1989     len = utf_ptr2len_len(p, maxlen);
1990     /* Only accept a composing char when the first char isn't illegal. */
1991     if ((len > 1 || *p < 0x80)
1992 	    && len < maxlen
1993 	    && p[len] >= 0x80
1994 	    && UTF_COMPOSINGLIKE(p, p + len))
1995     {
1996 	cc = utf_ptr2char(p + len);
1997 	for (;;)
1998 	{
1999 	    pcc[i++] = cc;
2000 	    if (i == MAX_MCO)
2001 		break;
2002 	    len += utf_ptr2len_len(p + len, maxlen - len);
2003 	    if (len >= maxlen
2004 		    || p[len] < 0x80
2005 		    || !utf_iscomposing(cc = utf_ptr2char(p + len)))
2006 		break;
2007 	}
2008     }
2009 
2010     if (i < MAX_MCO)	/* last composing char must be 0 */
2011 	pcc[i] = 0;
2012 
2013     return c;
2014 }
2015 
2016 /*
2017  * Convert the character at screen position "off" to a sequence of bytes.
2018  * Includes the composing characters.
2019  * "buf" must at least have the length MB_MAXBYTES + 1.
2020  * Only to be used when ScreenLinesUC[off] != 0.
2021  * Returns the produced number of bytes.
2022  */
2023     int
2024 utfc_char2bytes(int off, char_u *buf)
2025 {
2026     int		len;
2027     int		i;
2028 
2029     len = utf_char2bytes(ScreenLinesUC[off], buf);
2030     for (i = 0; i < Screen_mco; ++i)
2031     {
2032 	if (ScreenLinesC[i][off] == 0)
2033 	    break;
2034 	len += utf_char2bytes(ScreenLinesC[i][off], buf + len);
2035     }
2036     return len;
2037 }
2038 
2039 /*
2040  * Get the length of a UTF-8 byte sequence, not including any following
2041  * composing characters.
2042  * Returns 0 for "".
2043  * Returns 1 for an illegal byte sequence.
2044  */
2045     int
2046 utf_ptr2len(char_u *p)
2047 {
2048     int		len;
2049     int		i;
2050 
2051     if (*p == NUL)
2052 	return 0;
2053     len = utf8len_tab[*p];
2054     for (i = 1; i < len; ++i)
2055 	if ((p[i] & 0xc0) != 0x80)
2056 	    return 1;
2057     return len;
2058 }
2059 
2060 /*
2061  * Return length of UTF-8 character, obtained from the first byte.
2062  * "b" must be between 0 and 255!
2063  * Returns 1 for an invalid first byte value.
2064  */
2065     int
2066 utf_byte2len(int b)
2067 {
2068     return utf8len_tab[b];
2069 }
2070 
2071 /*
2072  * Get the length of UTF-8 byte sequence "p[size]".  Does not include any
2073  * following composing characters.
2074  * Returns 1 for "".
2075  * Returns 1 for an illegal byte sequence (also in incomplete byte seq.).
2076  * Returns number > "size" for an incomplete byte sequence.
2077  * Never returns zero.
2078  */
2079     int
2080 utf_ptr2len_len(char_u *p, int size)
2081 {
2082     int		len;
2083     int		i;
2084     int		m;
2085 
2086     len = utf8len_tab[*p];
2087     if (len == 1)
2088 	return 1;	/* NUL, ascii or illegal lead byte */
2089     if (len > size)
2090 	m = size;	/* incomplete byte sequence. */
2091     else
2092 	m = len;
2093     for (i = 1; i < m; ++i)
2094 	if ((p[i] & 0xc0) != 0x80)
2095 	    return 1;
2096     return len;
2097 }
2098 
2099 /*
2100  * Return the number of bytes the UTF-8 encoding of the character at "p" takes.
2101  * This includes following composing characters.
2102  */
2103     int
2104 utfc_ptr2len(char_u *p)
2105 {
2106     int		len;
2107     int		b0 = *p;
2108 #ifdef FEAT_ARABIC
2109     int		prevlen;
2110 #endif
2111 
2112     if (b0 == NUL)
2113 	return 0;
2114     if (b0 < 0x80 && p[1] < 0x80)	/* be quick for ASCII */
2115 	return 1;
2116 
2117     /* Skip over first UTF-8 char, stopping at a NUL byte. */
2118     len = utf_ptr2len(p);
2119 
2120     /* Check for illegal byte. */
2121     if (len == 1 && b0 >= 0x80)
2122 	return 1;
2123 
2124     /*
2125      * Check for composing characters.  We can handle only the first six, but
2126      * skip all of them (otherwise the cursor would get stuck).
2127      */
2128 #ifdef FEAT_ARABIC
2129     prevlen = 0;
2130 #endif
2131     for (;;)
2132     {
2133 	if (p[len] < 0x80 || !UTF_COMPOSINGLIKE(p + prevlen, p + len))
2134 	    return len;
2135 
2136 	/* Skip over composing char */
2137 #ifdef FEAT_ARABIC
2138 	prevlen = len;
2139 #endif
2140 	len += utf_ptr2len(p + len);
2141     }
2142 }
2143 
2144 /*
2145  * Return the number of bytes the UTF-8 encoding of the character at "p[size]"
2146  * takes.  This includes following composing characters.
2147  * Returns 0 for an empty string.
2148  * Returns 1 for an illegal char or an incomplete byte sequence.
2149  */
2150     int
2151 utfc_ptr2len_len(char_u *p, int size)
2152 {
2153     int		len;
2154 #ifdef FEAT_ARABIC
2155     int		prevlen;
2156 #endif
2157 
2158     if (size < 1 || *p == NUL)
2159 	return 0;
2160     if (p[0] < 0x80 && (size == 1 || p[1] < 0x80)) /* be quick for ASCII */
2161 	return 1;
2162 
2163     /* Skip over first UTF-8 char, stopping at a NUL byte. */
2164     len = utf_ptr2len_len(p, size);
2165 
2166     /* Check for illegal byte and incomplete byte sequence. */
2167     if ((len == 1 && p[0] >= 0x80) || len > size)
2168 	return 1;
2169 
2170     /*
2171      * Check for composing characters.  We can handle only the first six, but
2172      * skip all of them (otherwise the cursor would get stuck).
2173      */
2174 #ifdef FEAT_ARABIC
2175     prevlen = 0;
2176 #endif
2177     while (len < size)
2178     {
2179 	int	len_next_char;
2180 
2181 	if (p[len] < 0x80)
2182 	    break;
2183 
2184 	/*
2185 	 * Next character length should not go beyond size to ensure that
2186 	 * UTF_COMPOSINGLIKE(...) does not read beyond size.
2187 	 */
2188 	len_next_char = utf_ptr2len_len(p + len, size - len);
2189 	if (len_next_char > size - len)
2190 	    break;
2191 
2192 	if (!UTF_COMPOSINGLIKE(p + prevlen, p + len))
2193 	    break;
2194 
2195 	/* Skip over composing char */
2196 #ifdef FEAT_ARABIC
2197 	prevlen = len;
2198 #endif
2199 	len += len_next_char;
2200     }
2201     return len;
2202 }
2203 
2204 /*
2205  * Return the number of bytes the UTF-8 encoding of character "c" takes.
2206  * This does not include composing characters.
2207  */
2208     int
2209 utf_char2len(int c)
2210 {
2211     if (c < 0x80)
2212 	return 1;
2213     if (c < 0x800)
2214 	return 2;
2215     if (c < 0x10000)
2216 	return 3;
2217     if (c < 0x200000)
2218 	return 4;
2219     if (c < 0x4000000)
2220 	return 5;
2221     return 6;
2222 }
2223 
2224 /*
2225  * Convert Unicode character "c" to UTF-8 string in "buf[]".
2226  * Returns the number of bytes.
2227  */
2228     int
2229 utf_char2bytes(int c, char_u *buf)
2230 {
2231     if (c < 0x80)		/* 7 bits */
2232     {
2233 	buf[0] = c;
2234 	return 1;
2235     }
2236     if (c < 0x800)		/* 11 bits */
2237     {
2238 	buf[0] = 0xc0 + ((unsigned)c >> 6);
2239 	buf[1] = 0x80 + (c & 0x3f);
2240 	return 2;
2241     }
2242     if (c < 0x10000)		/* 16 bits */
2243     {
2244 	buf[0] = 0xe0 + ((unsigned)c >> 12);
2245 	buf[1] = 0x80 + (((unsigned)c >> 6) & 0x3f);
2246 	buf[2] = 0x80 + (c & 0x3f);
2247 	return 3;
2248     }
2249     if (c < 0x200000)		/* 21 bits */
2250     {
2251 	buf[0] = 0xf0 + ((unsigned)c >> 18);
2252 	buf[1] = 0x80 + (((unsigned)c >> 12) & 0x3f);
2253 	buf[2] = 0x80 + (((unsigned)c >> 6) & 0x3f);
2254 	buf[3] = 0x80 + (c & 0x3f);
2255 	return 4;
2256     }
2257     if (c < 0x4000000)		/* 26 bits */
2258     {
2259 	buf[0] = 0xf8 + ((unsigned)c >> 24);
2260 	buf[1] = 0x80 + (((unsigned)c >> 18) & 0x3f);
2261 	buf[2] = 0x80 + (((unsigned)c >> 12) & 0x3f);
2262 	buf[3] = 0x80 + (((unsigned)c >> 6) & 0x3f);
2263 	buf[4] = 0x80 + (c & 0x3f);
2264 	return 5;
2265     }
2266 				/* 31 bits */
2267     buf[0] = 0xfc + ((unsigned)c >> 30);
2268     buf[1] = 0x80 + (((unsigned)c >> 24) & 0x3f);
2269     buf[2] = 0x80 + (((unsigned)c >> 18) & 0x3f);
2270     buf[3] = 0x80 + (((unsigned)c >> 12) & 0x3f);
2271     buf[4] = 0x80 + (((unsigned)c >> 6) & 0x3f);
2272     buf[5] = 0x80 + (c & 0x3f);
2273     return 6;
2274 }
2275 
2276 #if defined(FEAT_TERMINAL) || defined(PROTO)
2277 /*
2278  * utf_iscomposing() with different argument type for libvterm.
2279  */
2280     int
2281 utf_iscomposing_uint(UINT32_T c)
2282 {
2283     return utf_iscomposing((int)c);
2284 }
2285 #endif
2286 
2287 /*
2288  * Return TRUE if "c" is a composing UTF-8 character.  This means it will be
2289  * drawn on top of the preceding character.
2290  * Based on code from Markus Kuhn.
2291  */
2292     int
2293 utf_iscomposing(int c)
2294 {
2295     /* Sorted list of non-overlapping intervals.
2296      * Generated by ../runtime/tools/unicode.vim. */
2297     static struct interval combining[] =
2298     {
2299 	{0x0300, 0x036f},
2300 	{0x0483, 0x0489},
2301 	{0x0591, 0x05bd},
2302 	{0x05bf, 0x05bf},
2303 	{0x05c1, 0x05c2},
2304 	{0x05c4, 0x05c5},
2305 	{0x05c7, 0x05c7},
2306 	{0x0610, 0x061a},
2307 	{0x064b, 0x065f},
2308 	{0x0670, 0x0670},
2309 	{0x06d6, 0x06dc},
2310 	{0x06df, 0x06e4},
2311 	{0x06e7, 0x06e8},
2312 	{0x06ea, 0x06ed},
2313 	{0x0711, 0x0711},
2314 	{0x0730, 0x074a},
2315 	{0x07a6, 0x07b0},
2316 	{0x07eb, 0x07f3},
2317 	{0x07fd, 0x07fd},
2318 	{0x0816, 0x0819},
2319 	{0x081b, 0x0823},
2320 	{0x0825, 0x0827},
2321 	{0x0829, 0x082d},
2322 	{0x0859, 0x085b},
2323 	{0x08d3, 0x08e1},
2324 	{0x08e3, 0x0903},
2325 	{0x093a, 0x093c},
2326 	{0x093e, 0x094f},
2327 	{0x0951, 0x0957},
2328 	{0x0962, 0x0963},
2329 	{0x0981, 0x0983},
2330 	{0x09bc, 0x09bc},
2331 	{0x09be, 0x09c4},
2332 	{0x09c7, 0x09c8},
2333 	{0x09cb, 0x09cd},
2334 	{0x09d7, 0x09d7},
2335 	{0x09e2, 0x09e3},
2336 	{0x09fe, 0x09fe},
2337 	{0x0a01, 0x0a03},
2338 	{0x0a3c, 0x0a3c},
2339 	{0x0a3e, 0x0a42},
2340 	{0x0a47, 0x0a48},
2341 	{0x0a4b, 0x0a4d},
2342 	{0x0a51, 0x0a51},
2343 	{0x0a70, 0x0a71},
2344 	{0x0a75, 0x0a75},
2345 	{0x0a81, 0x0a83},
2346 	{0x0abc, 0x0abc},
2347 	{0x0abe, 0x0ac5},
2348 	{0x0ac7, 0x0ac9},
2349 	{0x0acb, 0x0acd},
2350 	{0x0ae2, 0x0ae3},
2351 	{0x0afa, 0x0aff},
2352 	{0x0b01, 0x0b03},
2353 	{0x0b3c, 0x0b3c},
2354 	{0x0b3e, 0x0b44},
2355 	{0x0b47, 0x0b48},
2356 	{0x0b4b, 0x0b4d},
2357 	{0x0b56, 0x0b57},
2358 	{0x0b62, 0x0b63},
2359 	{0x0b82, 0x0b82},
2360 	{0x0bbe, 0x0bc2},
2361 	{0x0bc6, 0x0bc8},
2362 	{0x0bca, 0x0bcd},
2363 	{0x0bd7, 0x0bd7},
2364 	{0x0c00, 0x0c04},
2365 	{0x0c3e, 0x0c44},
2366 	{0x0c46, 0x0c48},
2367 	{0x0c4a, 0x0c4d},
2368 	{0x0c55, 0x0c56},
2369 	{0x0c62, 0x0c63},
2370 	{0x0c81, 0x0c83},
2371 	{0x0cbc, 0x0cbc},
2372 	{0x0cbe, 0x0cc4},
2373 	{0x0cc6, 0x0cc8},
2374 	{0x0cca, 0x0ccd},
2375 	{0x0cd5, 0x0cd6},
2376 	{0x0ce2, 0x0ce3},
2377 	{0x0d00, 0x0d03},
2378 	{0x0d3b, 0x0d3c},
2379 	{0x0d3e, 0x0d44},
2380 	{0x0d46, 0x0d48},
2381 	{0x0d4a, 0x0d4d},
2382 	{0x0d57, 0x0d57},
2383 	{0x0d62, 0x0d63},
2384 	{0x0d82, 0x0d83},
2385 	{0x0dca, 0x0dca},
2386 	{0x0dcf, 0x0dd4},
2387 	{0x0dd6, 0x0dd6},
2388 	{0x0dd8, 0x0ddf},
2389 	{0x0df2, 0x0df3},
2390 	{0x0e31, 0x0e31},
2391 	{0x0e34, 0x0e3a},
2392 	{0x0e47, 0x0e4e},
2393 	{0x0eb1, 0x0eb1},
2394 	{0x0eb4, 0x0eb9},
2395 	{0x0ebb, 0x0ebc},
2396 	{0x0ec8, 0x0ecd},
2397 	{0x0f18, 0x0f19},
2398 	{0x0f35, 0x0f35},
2399 	{0x0f37, 0x0f37},
2400 	{0x0f39, 0x0f39},
2401 	{0x0f3e, 0x0f3f},
2402 	{0x0f71, 0x0f84},
2403 	{0x0f86, 0x0f87},
2404 	{0x0f8d, 0x0f97},
2405 	{0x0f99, 0x0fbc},
2406 	{0x0fc6, 0x0fc6},
2407 	{0x102b, 0x103e},
2408 	{0x1056, 0x1059},
2409 	{0x105e, 0x1060},
2410 	{0x1062, 0x1064},
2411 	{0x1067, 0x106d},
2412 	{0x1071, 0x1074},
2413 	{0x1082, 0x108d},
2414 	{0x108f, 0x108f},
2415 	{0x109a, 0x109d},
2416 	{0x135d, 0x135f},
2417 	{0x1712, 0x1714},
2418 	{0x1732, 0x1734},
2419 	{0x1752, 0x1753},
2420 	{0x1772, 0x1773},
2421 	{0x17b4, 0x17d3},
2422 	{0x17dd, 0x17dd},
2423 	{0x180b, 0x180d},
2424 	{0x1885, 0x1886},
2425 	{0x18a9, 0x18a9},
2426 	{0x1920, 0x192b},
2427 	{0x1930, 0x193b},
2428 	{0x1a17, 0x1a1b},
2429 	{0x1a55, 0x1a5e},
2430 	{0x1a60, 0x1a7c},
2431 	{0x1a7f, 0x1a7f},
2432 	{0x1ab0, 0x1abe},
2433 	{0x1b00, 0x1b04},
2434 	{0x1b34, 0x1b44},
2435 	{0x1b6b, 0x1b73},
2436 	{0x1b80, 0x1b82},
2437 	{0x1ba1, 0x1bad},
2438 	{0x1be6, 0x1bf3},
2439 	{0x1c24, 0x1c37},
2440 	{0x1cd0, 0x1cd2},
2441 	{0x1cd4, 0x1ce8},
2442 	{0x1ced, 0x1ced},
2443 	{0x1cf2, 0x1cf4},
2444 	{0x1cf7, 0x1cf9},
2445 	{0x1dc0, 0x1df9},
2446 	{0x1dfb, 0x1dff},
2447 	{0x20d0, 0x20f0},
2448 	{0x2cef, 0x2cf1},
2449 	{0x2d7f, 0x2d7f},
2450 	{0x2de0, 0x2dff},
2451 	{0x302a, 0x302f},
2452 	{0x3099, 0x309a},
2453 	{0xa66f, 0xa672},
2454 	{0xa674, 0xa67d},
2455 	{0xa69e, 0xa69f},
2456 	{0xa6f0, 0xa6f1},
2457 	{0xa802, 0xa802},
2458 	{0xa806, 0xa806},
2459 	{0xa80b, 0xa80b},
2460 	{0xa823, 0xa827},
2461 	{0xa880, 0xa881},
2462 	{0xa8b4, 0xa8c5},
2463 	{0xa8e0, 0xa8f1},
2464 	{0xa8ff, 0xa8ff},
2465 	{0xa926, 0xa92d},
2466 	{0xa947, 0xa953},
2467 	{0xa980, 0xa983},
2468 	{0xa9b3, 0xa9c0},
2469 	{0xa9e5, 0xa9e5},
2470 	{0xaa29, 0xaa36},
2471 	{0xaa43, 0xaa43},
2472 	{0xaa4c, 0xaa4d},
2473 	{0xaa7b, 0xaa7d},
2474 	{0xaab0, 0xaab0},
2475 	{0xaab2, 0xaab4},
2476 	{0xaab7, 0xaab8},
2477 	{0xaabe, 0xaabf},
2478 	{0xaac1, 0xaac1},
2479 	{0xaaeb, 0xaaef},
2480 	{0xaaf5, 0xaaf6},
2481 	{0xabe3, 0xabea},
2482 	{0xabec, 0xabed},
2483 	{0xfb1e, 0xfb1e},
2484 	{0xfe00, 0xfe0f},
2485 	{0xfe20, 0xfe2f},
2486 	{0x101fd, 0x101fd},
2487 	{0x102e0, 0x102e0},
2488 	{0x10376, 0x1037a},
2489 	{0x10a01, 0x10a03},
2490 	{0x10a05, 0x10a06},
2491 	{0x10a0c, 0x10a0f},
2492 	{0x10a38, 0x10a3a},
2493 	{0x10a3f, 0x10a3f},
2494 	{0x10ae5, 0x10ae6},
2495 	{0x10d24, 0x10d27},
2496 	{0x10f46, 0x10f50},
2497 	{0x11000, 0x11002},
2498 	{0x11038, 0x11046},
2499 	{0x1107f, 0x11082},
2500 	{0x110b0, 0x110ba},
2501 	{0x11100, 0x11102},
2502 	{0x11127, 0x11134},
2503 	{0x11145, 0x11146},
2504 	{0x11173, 0x11173},
2505 	{0x11180, 0x11182},
2506 	{0x111b3, 0x111c0},
2507 	{0x111c9, 0x111cc},
2508 	{0x1122c, 0x11237},
2509 	{0x1123e, 0x1123e},
2510 	{0x112df, 0x112ea},
2511 	{0x11300, 0x11303},
2512 	{0x1133b, 0x1133c},
2513 	{0x1133e, 0x11344},
2514 	{0x11347, 0x11348},
2515 	{0x1134b, 0x1134d},
2516 	{0x11357, 0x11357},
2517 	{0x11362, 0x11363},
2518 	{0x11366, 0x1136c},
2519 	{0x11370, 0x11374},
2520 	{0x11435, 0x11446},
2521 	{0x1145e, 0x1145e},
2522 	{0x114b0, 0x114c3},
2523 	{0x115af, 0x115b5},
2524 	{0x115b8, 0x115c0},
2525 	{0x115dc, 0x115dd},
2526 	{0x11630, 0x11640},
2527 	{0x116ab, 0x116b7},
2528 	{0x1171d, 0x1172b},
2529 	{0x1182c, 0x1183a},
2530 	{0x11a01, 0x11a0a},
2531 	{0x11a33, 0x11a39},
2532 	{0x11a3b, 0x11a3e},
2533 	{0x11a47, 0x11a47},
2534 	{0x11a51, 0x11a5b},
2535 	{0x11a8a, 0x11a99},
2536 	{0x11c2f, 0x11c36},
2537 	{0x11c38, 0x11c3f},
2538 	{0x11c92, 0x11ca7},
2539 	{0x11ca9, 0x11cb6},
2540 	{0x11d31, 0x11d36},
2541 	{0x11d3a, 0x11d3a},
2542 	{0x11d3c, 0x11d3d},
2543 	{0x11d3f, 0x11d45},
2544 	{0x11d47, 0x11d47},
2545 	{0x11d8a, 0x11d8e},
2546 	{0x11d90, 0x11d91},
2547 	{0x11d93, 0x11d97},
2548 	{0x11ef3, 0x11ef6},
2549 	{0x16af0, 0x16af4},
2550 	{0x16b30, 0x16b36},
2551 	{0x16f51, 0x16f7e},
2552 	{0x16f8f, 0x16f92},
2553 	{0x1bc9d, 0x1bc9e},
2554 	{0x1d165, 0x1d169},
2555 	{0x1d16d, 0x1d172},
2556 	{0x1d17b, 0x1d182},
2557 	{0x1d185, 0x1d18b},
2558 	{0x1d1aa, 0x1d1ad},
2559 	{0x1d242, 0x1d244},
2560 	{0x1da00, 0x1da36},
2561 	{0x1da3b, 0x1da6c},
2562 	{0x1da75, 0x1da75},
2563 	{0x1da84, 0x1da84},
2564 	{0x1da9b, 0x1da9f},
2565 	{0x1daa1, 0x1daaf},
2566 	{0x1e000, 0x1e006},
2567 	{0x1e008, 0x1e018},
2568 	{0x1e01b, 0x1e021},
2569 	{0x1e023, 0x1e024},
2570 	{0x1e026, 0x1e02a},
2571 	{0x1e8d0, 0x1e8d6},
2572 	{0x1e944, 0x1e94a},
2573 	{0xe0100, 0xe01ef}
2574     };
2575 
2576     return intable(combining, sizeof(combining), c);
2577 }
2578 
2579 /*
2580  * Return TRUE for characters that can be displayed in a normal way.
2581  * Only for characters of 0x100 and above!
2582  */
2583     int
2584 utf_printable(int c)
2585 {
2586 #ifdef USE_WCHAR_FUNCTIONS
2587     /*
2588      * Assume the iswprint() library function works better than our own stuff.
2589      */
2590     return iswprint(c);
2591 #else
2592     /* Sorted list of non-overlapping intervals.
2593      * 0xd800-0xdfff is reserved for UTF-16, actually illegal. */
2594     static struct interval nonprint[] =
2595     {
2596 	{0x070f, 0x070f}, {0x180b, 0x180e}, {0x200b, 0x200f}, {0x202a, 0x202e},
2597 	{0x206a, 0x206f}, {0xd800, 0xdfff}, {0xfeff, 0xfeff}, {0xfff9, 0xfffb},
2598 	{0xfffe, 0xffff}
2599     };
2600 
2601     return !intable(nonprint, sizeof(nonprint), c);
2602 #endif
2603 }
2604 
2605 /* Sorted list of non-overlapping intervals of all Emoji characters,
2606  * based on http://unicode.org/emoji/charts/emoji-list.html */
2607 static struct interval emoji_all[] =
2608 {
2609     {0x203c, 0x203c},
2610     {0x2049, 0x2049},
2611     {0x2122, 0x2122},
2612     {0x2139, 0x2139},
2613     {0x2194, 0x2199},
2614     {0x21a9, 0x21aa},
2615     {0x231a, 0x231b},
2616     {0x2328, 0x2328},
2617     {0x23cf, 0x23cf},
2618     {0x23e9, 0x23f3},
2619     {0x23f8, 0x23fa},
2620     {0x24c2, 0x24c2},
2621     {0x25aa, 0x25ab},
2622     {0x25b6, 0x25b6},
2623     {0x25c0, 0x25c0},
2624     {0x25fb, 0x25fe},
2625     {0x2600, 0x2604},
2626     {0x260e, 0x260e},
2627     {0x2611, 0x2611},
2628     {0x2614, 0x2615},
2629     {0x2618, 0x2618},
2630     {0x261d, 0x261d},
2631     {0x2620, 0x2620},
2632     {0x2622, 0x2623},
2633     {0x2626, 0x2626},
2634     {0x262a, 0x262a},
2635     {0x262e, 0x262f},
2636     {0x2638, 0x263a},
2637     {0x2640, 0x2640},
2638     {0x2642, 0x2642},
2639     {0x2648, 0x2653},
2640     {0x265f, 0x2660},
2641     {0x2663, 0x2663},
2642     {0x2665, 0x2666},
2643     {0x2668, 0x2668},
2644     {0x267b, 0x267b},
2645     {0x267e, 0x267f},
2646     {0x2692, 0x2697},
2647     {0x2699, 0x2699},
2648     {0x269b, 0x269c},
2649     {0x26a0, 0x26a1},
2650     {0x26aa, 0x26ab},
2651     {0x26b0, 0x26b1},
2652     {0x26bd, 0x26be},
2653     {0x26c4, 0x26c5},
2654     {0x26c8, 0x26c8},
2655     {0x26ce, 0x26cf},
2656     {0x26d1, 0x26d1},
2657     {0x26d3, 0x26d4},
2658     {0x26e9, 0x26ea},
2659     {0x26f0, 0x26f5},
2660     {0x26f7, 0x26fa},
2661     {0x26fd, 0x26fd},
2662     {0x2702, 0x2702},
2663     {0x2705, 0x2705},
2664     {0x2708, 0x270d},
2665     {0x270f, 0x270f},
2666     {0x2712, 0x2712},
2667     {0x2714, 0x2714},
2668     {0x2716, 0x2716},
2669     {0x271d, 0x271d},
2670     {0x2721, 0x2721},
2671     {0x2728, 0x2728},
2672     {0x2733, 0x2734},
2673     {0x2744, 0x2744},
2674     {0x2747, 0x2747},
2675     {0x274c, 0x274c},
2676     {0x274e, 0x274e},
2677     {0x2753, 0x2755},
2678     {0x2757, 0x2757},
2679     {0x2763, 0x2764},
2680     {0x2795, 0x2797},
2681     {0x27a1, 0x27a1},
2682     {0x27b0, 0x27b0},
2683     {0x27bf, 0x27bf},
2684     {0x2934, 0x2935},
2685     {0x2b05, 0x2b07},
2686     {0x2b1b, 0x2b1c},
2687     {0x2b50, 0x2b50},
2688     {0x2b55, 0x2b55},
2689     {0x3030, 0x3030},
2690     {0x303d, 0x303d},
2691     {0x3297, 0x3297},
2692     {0x3299, 0x3299},
2693     {0x1f004, 0x1f004},
2694     {0x1f0cf, 0x1f0cf},
2695     {0x1f170, 0x1f171},
2696     {0x1f17e, 0x1f17f},
2697     {0x1f18e, 0x1f18e},
2698     {0x1f191, 0x1f19a},
2699     {0x1f1e6, 0x1f1ff},
2700     {0x1f201, 0x1f202},
2701     {0x1f21a, 0x1f21a},
2702     {0x1f22f, 0x1f22f},
2703     {0x1f232, 0x1f23a},
2704     {0x1f250, 0x1f251},
2705     {0x1f300, 0x1f321},
2706     {0x1f324, 0x1f393},
2707     {0x1f396, 0x1f397},
2708     {0x1f399, 0x1f39b},
2709     {0x1f39e, 0x1f3f0},
2710     {0x1f3f3, 0x1f3f5},
2711     {0x1f3f7, 0x1f4fd},
2712     {0x1f4ff, 0x1f53d},
2713     {0x1f549, 0x1f54e},
2714     {0x1f550, 0x1f567},
2715     {0x1f56f, 0x1f570},
2716     {0x1f573, 0x1f57a},
2717     {0x1f587, 0x1f587},
2718     {0x1f58a, 0x1f58d},
2719     {0x1f590, 0x1f590},
2720     {0x1f595, 0x1f596},
2721     {0x1f5a4, 0x1f5a5},
2722     {0x1f5a8, 0x1f5a8},
2723     {0x1f5b1, 0x1f5b2},
2724     {0x1f5bc, 0x1f5bc},
2725     {0x1f5c2, 0x1f5c4},
2726     {0x1f5d1, 0x1f5d3},
2727     {0x1f5dc, 0x1f5de},
2728     {0x1f5e1, 0x1f5e1},
2729     {0x1f5e3, 0x1f5e3},
2730     {0x1f5e8, 0x1f5e8},
2731     {0x1f5ef, 0x1f5ef},
2732     {0x1f5f3, 0x1f5f3},
2733     {0x1f5fa, 0x1f64f},
2734     {0x1f680, 0x1f6c5},
2735     {0x1f6cb, 0x1f6d2},
2736     {0x1f6e0, 0x1f6e5},
2737     {0x1f6e9, 0x1f6e9},
2738     {0x1f6eb, 0x1f6ec},
2739     {0x1f6f0, 0x1f6f0},
2740     {0x1f6f3, 0x1f6f9},
2741     {0x1f910, 0x1f93a},
2742     {0x1f93c, 0x1f93e},
2743     {0x1f940, 0x1f945},
2744     {0x1f947, 0x1f970},
2745     {0x1f973, 0x1f976},
2746     {0x1f97a, 0x1f97a},
2747     {0x1f97c, 0x1f9a2},
2748     {0x1f9b0, 0x1f9b9},
2749     {0x1f9c0, 0x1f9c2},
2750     {0x1f9d0, 0x1f9ff}
2751 };
2752 
2753 /*
2754  * Get class of a Unicode character.
2755  * 0: white space
2756  * 1: punctuation
2757  * 2 or bigger: some class of word character.
2758  */
2759     int
2760 utf_class(int c)
2761 {
2762     return utf_class_buf(c, curbuf);
2763 }
2764 
2765     int
2766 utf_class_buf(int c, buf_T *buf)
2767 {
2768     /* sorted list of non-overlapping intervals */
2769     static struct clinterval
2770     {
2771 	unsigned int first;
2772 	unsigned int last;
2773 	unsigned int class;
2774     } classes[] =
2775     {
2776 	{0x037e, 0x037e, 1},		/* Greek question mark */
2777 	{0x0387, 0x0387, 1},		/* Greek ano teleia */
2778 	{0x055a, 0x055f, 1},		/* Armenian punctuation */
2779 	{0x0589, 0x0589, 1},		/* Armenian full stop */
2780 	{0x05be, 0x05be, 1},
2781 	{0x05c0, 0x05c0, 1},
2782 	{0x05c3, 0x05c3, 1},
2783 	{0x05f3, 0x05f4, 1},
2784 	{0x060c, 0x060c, 1},
2785 	{0x061b, 0x061b, 1},
2786 	{0x061f, 0x061f, 1},
2787 	{0x066a, 0x066d, 1},
2788 	{0x06d4, 0x06d4, 1},
2789 	{0x0700, 0x070d, 1},		/* Syriac punctuation */
2790 	{0x0964, 0x0965, 1},
2791 	{0x0970, 0x0970, 1},
2792 	{0x0df4, 0x0df4, 1},
2793 	{0x0e4f, 0x0e4f, 1},
2794 	{0x0e5a, 0x0e5b, 1},
2795 	{0x0f04, 0x0f12, 1},
2796 	{0x0f3a, 0x0f3d, 1},
2797 	{0x0f85, 0x0f85, 1},
2798 	{0x104a, 0x104f, 1},		/* Myanmar punctuation */
2799 	{0x10fb, 0x10fb, 1},		/* Georgian punctuation */
2800 	{0x1361, 0x1368, 1},		/* Ethiopic punctuation */
2801 	{0x166d, 0x166e, 1},		/* Canadian Syl. punctuation */
2802 	{0x1680, 0x1680, 0},
2803 	{0x169b, 0x169c, 1},
2804 	{0x16eb, 0x16ed, 1},
2805 	{0x1735, 0x1736, 1},
2806 	{0x17d4, 0x17dc, 1},		/* Khmer punctuation */
2807 	{0x1800, 0x180a, 1},		/* Mongolian punctuation */
2808 	{0x2000, 0x200b, 0},		/* spaces */
2809 	{0x200c, 0x2027, 1},		/* punctuation and symbols */
2810 	{0x2028, 0x2029, 0},
2811 	{0x202a, 0x202e, 1},		/* punctuation and symbols */
2812 	{0x202f, 0x202f, 0},
2813 	{0x2030, 0x205e, 1},		/* punctuation and symbols */
2814 	{0x205f, 0x205f, 0},
2815 	{0x2060, 0x27ff, 1},		/* punctuation and symbols */
2816 	{0x2070, 0x207f, 0x2070},	/* superscript */
2817 	{0x2080, 0x2094, 0x2080},	/* subscript */
2818 	{0x20a0, 0x27ff, 1},		/* all kinds of symbols */
2819 	{0x2800, 0x28ff, 0x2800},	/* braille */
2820 	{0x2900, 0x2998, 1},		/* arrows, brackets, etc. */
2821 	{0x29d8, 0x29db, 1},
2822 	{0x29fc, 0x29fd, 1},
2823 	{0x2e00, 0x2e7f, 1},		/* supplemental punctuation */
2824 	{0x3000, 0x3000, 0},		/* ideographic space */
2825 	{0x3001, 0x3020, 1},		/* ideographic punctuation */
2826 	{0x3030, 0x3030, 1},
2827 	{0x303d, 0x303d, 1},
2828 	{0x3040, 0x309f, 0x3040},	/* Hiragana */
2829 	{0x30a0, 0x30ff, 0x30a0},	/* Katakana */
2830 	{0x3300, 0x9fff, 0x4e00},	/* CJK Ideographs */
2831 	{0xac00, 0xd7a3, 0xac00},	/* Hangul Syllables */
2832 	{0xf900, 0xfaff, 0x4e00},	/* CJK Ideographs */
2833 	{0xfd3e, 0xfd3f, 1},
2834 	{0xfe30, 0xfe6b, 1},		/* punctuation forms */
2835 	{0xff00, 0xff0f, 1},		/* half/fullwidth ASCII */
2836 	{0xff1a, 0xff20, 1},		/* half/fullwidth ASCII */
2837 	{0xff3b, 0xff40, 1},		/* half/fullwidth ASCII */
2838 	{0xff5b, 0xff65, 1},		/* half/fullwidth ASCII */
2839 	{0x20000, 0x2a6df, 0x4e00},	/* CJK Ideographs */
2840 	{0x2a700, 0x2b73f, 0x4e00},	/* CJK Ideographs */
2841 	{0x2b740, 0x2b81f, 0x4e00},	/* CJK Ideographs */
2842 	{0x2f800, 0x2fa1f, 0x4e00},	/* CJK Ideographs */
2843     };
2844 
2845     int bot = 0;
2846     int top = sizeof(classes) / sizeof(struct clinterval) - 1;
2847     int mid;
2848 
2849     /* First quick check for Latin1 characters, use 'iskeyword'. */
2850     if (c < 0x100)
2851     {
2852 	if (c == ' ' || c == '\t' || c == NUL || c == 0xa0)
2853 	    return 0;	    /* blank */
2854 	if (vim_iswordc_buf(c, buf))
2855 	    return 2;	    /* word character */
2856 	return 1;	    /* punctuation */
2857     }
2858 
2859     /* binary search in table */
2860     while (top >= bot)
2861     {
2862 	mid = (bot + top) / 2;
2863 	if (classes[mid].last < (unsigned int)c)
2864 	    bot = mid + 1;
2865 	else if (classes[mid].first > (unsigned int)c)
2866 	    top = mid - 1;
2867 	else
2868 	    return (int)classes[mid].class;
2869     }
2870 
2871     /* emoji */
2872     if (intable(emoji_all, sizeof(emoji_all), c))
2873 	return 3;
2874 
2875     /* most other characters are "word" characters */
2876     return 2;
2877 }
2878 
2879     int
2880 utf_ambiguous_width(int c)
2881 {
2882     return c >= 0x80 && (intable(ambiguous, sizeof(ambiguous), c)
2883 	    || intable(emoji_all, sizeof(emoji_all), c));
2884 }
2885 
2886 /*
2887  * Code for Unicode case-dependent operations.  Based on notes in
2888  * http://www.unicode.org/Public/UNIDATA/CaseFolding.txt
2889  * This code uses simple case folding, not full case folding.
2890  * Last updated for Unicode 5.2.
2891  */
2892 
2893 /*
2894  * The following tables are built by ../runtime/tools/unicode.vim.
2895  * They must be in numeric order, because we use binary search.
2896  * An entry such as {0x41,0x5a,1,32} means that Unicode characters in the
2897  * range from 0x41 to 0x5a inclusive, stepping by 1, are changed to
2898  * folded/upper/lower by adding 32.
2899  */
2900 typedef struct
2901 {
2902     int rangeStart;
2903     int rangeEnd;
2904     int step;
2905     int offset;
2906 } convertStruct;
2907 
2908 static convertStruct foldCase[] =
2909 {
2910 	{0x41,0x5a,1,32},
2911 	{0xb5,0xb5,-1,775},
2912 	{0xc0,0xd6,1,32},
2913 	{0xd8,0xde,1,32},
2914 	{0x100,0x12e,2,1},
2915 	{0x132,0x136,2,1},
2916 	{0x139,0x147,2,1},
2917 	{0x14a,0x176,2,1},
2918 	{0x178,0x178,-1,-121},
2919 	{0x179,0x17d,2,1},
2920 	{0x17f,0x17f,-1,-268},
2921 	{0x181,0x181,-1,210},
2922 	{0x182,0x184,2,1},
2923 	{0x186,0x186,-1,206},
2924 	{0x187,0x187,-1,1},
2925 	{0x189,0x18a,1,205},
2926 	{0x18b,0x18b,-1,1},
2927 	{0x18e,0x18e,-1,79},
2928 	{0x18f,0x18f,-1,202},
2929 	{0x190,0x190,-1,203},
2930 	{0x191,0x191,-1,1},
2931 	{0x193,0x193,-1,205},
2932 	{0x194,0x194,-1,207},
2933 	{0x196,0x196,-1,211},
2934 	{0x197,0x197,-1,209},
2935 	{0x198,0x198,-1,1},
2936 	{0x19c,0x19c,-1,211},
2937 	{0x19d,0x19d,-1,213},
2938 	{0x19f,0x19f,-1,214},
2939 	{0x1a0,0x1a4,2,1},
2940 	{0x1a6,0x1a6,-1,218},
2941 	{0x1a7,0x1a7,-1,1},
2942 	{0x1a9,0x1a9,-1,218},
2943 	{0x1ac,0x1ac,-1,1},
2944 	{0x1ae,0x1ae,-1,218},
2945 	{0x1af,0x1af,-1,1},
2946 	{0x1b1,0x1b2,1,217},
2947 	{0x1b3,0x1b5,2,1},
2948 	{0x1b7,0x1b7,-1,219},
2949 	{0x1b8,0x1bc,4,1},
2950 	{0x1c4,0x1c4,-1,2},
2951 	{0x1c5,0x1c5,-1,1},
2952 	{0x1c7,0x1c7,-1,2},
2953 	{0x1c8,0x1c8,-1,1},
2954 	{0x1ca,0x1ca,-1,2},
2955 	{0x1cb,0x1db,2,1},
2956 	{0x1de,0x1ee,2,1},
2957 	{0x1f1,0x1f1,-1,2},
2958 	{0x1f2,0x1f4,2,1},
2959 	{0x1f6,0x1f6,-1,-97},
2960 	{0x1f7,0x1f7,-1,-56},
2961 	{0x1f8,0x21e,2,1},
2962 	{0x220,0x220,-1,-130},
2963 	{0x222,0x232,2,1},
2964 	{0x23a,0x23a,-1,10795},
2965 	{0x23b,0x23b,-1,1},
2966 	{0x23d,0x23d,-1,-163},
2967 	{0x23e,0x23e,-1,10792},
2968 	{0x241,0x241,-1,1},
2969 	{0x243,0x243,-1,-195},
2970 	{0x244,0x244,-1,69},
2971 	{0x245,0x245,-1,71},
2972 	{0x246,0x24e,2,1},
2973 	{0x345,0x345,-1,116},
2974 	{0x370,0x372,2,1},
2975 	{0x376,0x376,-1,1},
2976 	{0x37f,0x37f,-1,116},
2977 	{0x386,0x386,-1,38},
2978 	{0x388,0x38a,1,37},
2979 	{0x38c,0x38c,-1,64},
2980 	{0x38e,0x38f,1,63},
2981 	{0x391,0x3a1,1,32},
2982 	{0x3a3,0x3ab,1,32},
2983 	{0x3c2,0x3c2,-1,1},
2984 	{0x3cf,0x3cf,-1,8},
2985 	{0x3d0,0x3d0,-1,-30},
2986 	{0x3d1,0x3d1,-1,-25},
2987 	{0x3d5,0x3d5,-1,-15},
2988 	{0x3d6,0x3d6,-1,-22},
2989 	{0x3d8,0x3ee,2,1},
2990 	{0x3f0,0x3f0,-1,-54},
2991 	{0x3f1,0x3f1,-1,-48},
2992 	{0x3f4,0x3f4,-1,-60},
2993 	{0x3f5,0x3f5,-1,-64},
2994 	{0x3f7,0x3f7,-1,1},
2995 	{0x3f9,0x3f9,-1,-7},
2996 	{0x3fa,0x3fa,-1,1},
2997 	{0x3fd,0x3ff,1,-130},
2998 	{0x400,0x40f,1,80},
2999 	{0x410,0x42f,1,32},
3000 	{0x460,0x480,2,1},
3001 	{0x48a,0x4be,2,1},
3002 	{0x4c0,0x4c0,-1,15},
3003 	{0x4c1,0x4cd,2,1},
3004 	{0x4d0,0x52e,2,1},
3005 	{0x531,0x556,1,48},
3006 	{0x10a0,0x10c5,1,7264},
3007 	{0x10c7,0x10cd,6,7264},
3008 	{0x13f8,0x13fd,1,-8},
3009 	{0x1c80,0x1c80,-1,-6222},
3010 	{0x1c81,0x1c81,-1,-6221},
3011 	{0x1c82,0x1c82,-1,-6212},
3012 	{0x1c83,0x1c84,1,-6210},
3013 	{0x1c85,0x1c85,-1,-6211},
3014 	{0x1c86,0x1c86,-1,-6204},
3015 	{0x1c87,0x1c87,-1,-6180},
3016 	{0x1c88,0x1c88,-1,35267},
3017 	{0x1c90,0x1cba,1,-3008},
3018 	{0x1cbd,0x1cbf,1,-3008},
3019 	{0x1e00,0x1e94,2,1},
3020 	{0x1e9b,0x1e9b,-1,-58},
3021 	{0x1e9e,0x1e9e,-1,-7615},
3022 	{0x1ea0,0x1efe,2,1},
3023 	{0x1f08,0x1f0f,1,-8},
3024 	{0x1f18,0x1f1d,1,-8},
3025 	{0x1f28,0x1f2f,1,-8},
3026 	{0x1f38,0x1f3f,1,-8},
3027 	{0x1f48,0x1f4d,1,-8},
3028 	{0x1f59,0x1f5f,2,-8},
3029 	{0x1f68,0x1f6f,1,-8},
3030 	{0x1f88,0x1f8f,1,-8},
3031 	{0x1f98,0x1f9f,1,-8},
3032 	{0x1fa8,0x1faf,1,-8},
3033 	{0x1fb8,0x1fb9,1,-8},
3034 	{0x1fba,0x1fbb,1,-74},
3035 	{0x1fbc,0x1fbc,-1,-9},
3036 	{0x1fbe,0x1fbe,-1,-7173},
3037 	{0x1fc8,0x1fcb,1,-86},
3038 	{0x1fcc,0x1fcc,-1,-9},
3039 	{0x1fd8,0x1fd9,1,-8},
3040 	{0x1fda,0x1fdb,1,-100},
3041 	{0x1fe8,0x1fe9,1,-8},
3042 	{0x1fea,0x1feb,1,-112},
3043 	{0x1fec,0x1fec,-1,-7},
3044 	{0x1ff8,0x1ff9,1,-128},
3045 	{0x1ffa,0x1ffb,1,-126},
3046 	{0x1ffc,0x1ffc,-1,-9},
3047 	{0x2126,0x2126,-1,-7517},
3048 	{0x212a,0x212a,-1,-8383},
3049 	{0x212b,0x212b,-1,-8262},
3050 	{0x2132,0x2132,-1,28},
3051 	{0x2160,0x216f,1,16},
3052 	{0x2183,0x2183,-1,1},
3053 	{0x24b6,0x24cf,1,26},
3054 	{0x2c00,0x2c2e,1,48},
3055 	{0x2c60,0x2c60,-1,1},
3056 	{0x2c62,0x2c62,-1,-10743},
3057 	{0x2c63,0x2c63,-1,-3814},
3058 	{0x2c64,0x2c64,-1,-10727},
3059 	{0x2c67,0x2c6b,2,1},
3060 	{0x2c6d,0x2c6d,-1,-10780},
3061 	{0x2c6e,0x2c6e,-1,-10749},
3062 	{0x2c6f,0x2c6f,-1,-10783},
3063 	{0x2c70,0x2c70,-1,-10782},
3064 	{0x2c72,0x2c75,3,1},
3065 	{0x2c7e,0x2c7f,1,-10815},
3066 	{0x2c80,0x2ce2,2,1},
3067 	{0x2ceb,0x2ced,2,1},
3068 	{0x2cf2,0xa640,31054,1},
3069 	{0xa642,0xa66c,2,1},
3070 	{0xa680,0xa69a,2,1},
3071 	{0xa722,0xa72e,2,1},
3072 	{0xa732,0xa76e,2,1},
3073 	{0xa779,0xa77b,2,1},
3074 	{0xa77d,0xa77d,-1,-35332},
3075 	{0xa77e,0xa786,2,1},
3076 	{0xa78b,0xa78b,-1,1},
3077 	{0xa78d,0xa78d,-1,-42280},
3078 	{0xa790,0xa792,2,1},
3079 	{0xa796,0xa7a8,2,1},
3080 	{0xa7aa,0xa7aa,-1,-42308},
3081 	{0xa7ab,0xa7ab,-1,-42319},
3082 	{0xa7ac,0xa7ac,-1,-42315},
3083 	{0xa7ad,0xa7ad,-1,-42305},
3084 	{0xa7ae,0xa7ae,-1,-42308},
3085 	{0xa7b0,0xa7b0,-1,-42258},
3086 	{0xa7b1,0xa7b1,-1,-42282},
3087 	{0xa7b2,0xa7b2,-1,-42261},
3088 	{0xa7b3,0xa7b3,-1,928},
3089 	{0xa7b4,0xa7b8,2,1},
3090 	{0xab70,0xabbf,1,-38864},
3091 	{0xff21,0xff3a,1,32},
3092 	{0x10400,0x10427,1,40},
3093 	{0x104b0,0x104d3,1,40},
3094 	{0x10c80,0x10cb2,1,64},
3095 	{0x118a0,0x118bf,1,32},
3096 	{0x16e40,0x16e5f,1,32},
3097 	{0x1e900,0x1e921,1,34}
3098 };
3099 
3100 /*
3101  * Generic conversion function for case operations.
3102  * Return the converted equivalent of "a", which is a UCS-4 character.  Use
3103  * the given conversion "table".  Uses binary search on "table".
3104  */
3105     static int
3106 utf_convert(
3107     int			a,
3108     convertStruct	table[],
3109     int			tableSize)
3110 {
3111     int start, mid, end; /* indices into table */
3112     int entries = tableSize / sizeof(convertStruct);
3113 
3114     start = 0;
3115     end = entries;
3116     while (start < end)
3117     {
3118 	/* need to search further */
3119 	mid = (end + start) / 2;
3120 	if (table[mid].rangeEnd < a)
3121 	    start = mid + 1;
3122 	else
3123 	    end = mid;
3124     }
3125     if (start < entries
3126 	    && table[start].rangeStart <= a
3127 	    && a <= table[start].rangeEnd
3128 	    && (a - table[start].rangeStart) % table[start].step == 0)
3129 	return (a + table[start].offset);
3130     else
3131 	return a;
3132 }
3133 
3134 /*
3135  * Return the folded-case equivalent of "a", which is a UCS-4 character.  Uses
3136  * simple case folding.
3137  */
3138     int
3139 utf_fold(int a)
3140 {
3141     if (a < 0x80)
3142 	/* be fast for ASCII */
3143 	return a >= 0x41 && a <= 0x5a ? a + 32 : a;
3144     return utf_convert(a, foldCase, (int)sizeof(foldCase));
3145 }
3146 
3147 static convertStruct toLower[] =
3148 {
3149 	{0x41,0x5a,1,32},
3150 	{0xc0,0xd6,1,32},
3151 	{0xd8,0xde,1,32},
3152 	{0x100,0x12e,2,1},
3153 	{0x130,0x130,-1,-199},
3154 	{0x132,0x136,2,1},
3155 	{0x139,0x147,2,1},
3156 	{0x14a,0x176,2,1},
3157 	{0x178,0x178,-1,-121},
3158 	{0x179,0x17d,2,1},
3159 	{0x181,0x181,-1,210},
3160 	{0x182,0x184,2,1},
3161 	{0x186,0x186,-1,206},
3162 	{0x187,0x187,-1,1},
3163 	{0x189,0x18a,1,205},
3164 	{0x18b,0x18b,-1,1},
3165 	{0x18e,0x18e,-1,79},
3166 	{0x18f,0x18f,-1,202},
3167 	{0x190,0x190,-1,203},
3168 	{0x191,0x191,-1,1},
3169 	{0x193,0x193,-1,205},
3170 	{0x194,0x194,-1,207},
3171 	{0x196,0x196,-1,211},
3172 	{0x197,0x197,-1,209},
3173 	{0x198,0x198,-1,1},
3174 	{0x19c,0x19c,-1,211},
3175 	{0x19d,0x19d,-1,213},
3176 	{0x19f,0x19f,-1,214},
3177 	{0x1a0,0x1a4,2,1},
3178 	{0x1a6,0x1a6,-1,218},
3179 	{0x1a7,0x1a7,-1,1},
3180 	{0x1a9,0x1a9,-1,218},
3181 	{0x1ac,0x1ac,-1,1},
3182 	{0x1ae,0x1ae,-1,218},
3183 	{0x1af,0x1af,-1,1},
3184 	{0x1b1,0x1b2,1,217},
3185 	{0x1b3,0x1b5,2,1},
3186 	{0x1b7,0x1b7,-1,219},
3187 	{0x1b8,0x1bc,4,1},
3188 	{0x1c4,0x1c4,-1,2},
3189 	{0x1c5,0x1c5,-1,1},
3190 	{0x1c7,0x1c7,-1,2},
3191 	{0x1c8,0x1c8,-1,1},
3192 	{0x1ca,0x1ca,-1,2},
3193 	{0x1cb,0x1db,2,1},
3194 	{0x1de,0x1ee,2,1},
3195 	{0x1f1,0x1f1,-1,2},
3196 	{0x1f2,0x1f4,2,1},
3197 	{0x1f6,0x1f6,-1,-97},
3198 	{0x1f7,0x1f7,-1,-56},
3199 	{0x1f8,0x21e,2,1},
3200 	{0x220,0x220,-1,-130},
3201 	{0x222,0x232,2,1},
3202 	{0x23a,0x23a,-1,10795},
3203 	{0x23b,0x23b,-1,1},
3204 	{0x23d,0x23d,-1,-163},
3205 	{0x23e,0x23e,-1,10792},
3206 	{0x241,0x241,-1,1},
3207 	{0x243,0x243,-1,-195},
3208 	{0x244,0x244,-1,69},
3209 	{0x245,0x245,-1,71},
3210 	{0x246,0x24e,2,1},
3211 	{0x370,0x372,2,1},
3212 	{0x376,0x376,-1,1},
3213 	{0x37f,0x37f,-1,116},
3214 	{0x386,0x386,-1,38},
3215 	{0x388,0x38a,1,37},
3216 	{0x38c,0x38c,-1,64},
3217 	{0x38e,0x38f,1,63},
3218 	{0x391,0x3a1,1,32},
3219 	{0x3a3,0x3ab,1,32},
3220 	{0x3cf,0x3cf,-1,8},
3221 	{0x3d8,0x3ee,2,1},
3222 	{0x3f4,0x3f4,-1,-60},
3223 	{0x3f7,0x3f7,-1,1},
3224 	{0x3f9,0x3f9,-1,-7},
3225 	{0x3fa,0x3fa,-1,1},
3226 	{0x3fd,0x3ff,1,-130},
3227 	{0x400,0x40f,1,80},
3228 	{0x410,0x42f,1,32},
3229 	{0x460,0x480,2,1},
3230 	{0x48a,0x4be,2,1},
3231 	{0x4c0,0x4c0,-1,15},
3232 	{0x4c1,0x4cd,2,1},
3233 	{0x4d0,0x52e,2,1},
3234 	{0x531,0x556,1,48},
3235 	{0x10a0,0x10c5,1,7264},
3236 	{0x10c7,0x10cd,6,7264},
3237 	{0x13a0,0x13ef,1,38864},
3238 	{0x13f0,0x13f5,1,8},
3239 	{0x1c90,0x1cba,1,-3008},
3240 	{0x1cbd,0x1cbf,1,-3008},
3241 	{0x1e00,0x1e94,2,1},
3242 	{0x1e9e,0x1e9e,-1,-7615},
3243 	{0x1ea0,0x1efe,2,1},
3244 	{0x1f08,0x1f0f,1,-8},
3245 	{0x1f18,0x1f1d,1,-8},
3246 	{0x1f28,0x1f2f,1,-8},
3247 	{0x1f38,0x1f3f,1,-8},
3248 	{0x1f48,0x1f4d,1,-8},
3249 	{0x1f59,0x1f5f,2,-8},
3250 	{0x1f68,0x1f6f,1,-8},
3251 	{0x1f88,0x1f8f,1,-8},
3252 	{0x1f98,0x1f9f,1,-8},
3253 	{0x1fa8,0x1faf,1,-8},
3254 	{0x1fb8,0x1fb9,1,-8},
3255 	{0x1fba,0x1fbb,1,-74},
3256 	{0x1fbc,0x1fbc,-1,-9},
3257 	{0x1fc8,0x1fcb,1,-86},
3258 	{0x1fcc,0x1fcc,-1,-9},
3259 	{0x1fd8,0x1fd9,1,-8},
3260 	{0x1fda,0x1fdb,1,-100},
3261 	{0x1fe8,0x1fe9,1,-8},
3262 	{0x1fea,0x1feb,1,-112},
3263 	{0x1fec,0x1fec,-1,-7},
3264 	{0x1ff8,0x1ff9,1,-128},
3265 	{0x1ffa,0x1ffb,1,-126},
3266 	{0x1ffc,0x1ffc,-1,-9},
3267 	{0x2126,0x2126,-1,-7517},
3268 	{0x212a,0x212a,-1,-8383},
3269 	{0x212b,0x212b,-1,-8262},
3270 	{0x2132,0x2132,-1,28},
3271 	{0x2160,0x216f,1,16},
3272 	{0x2183,0x2183,-1,1},
3273 	{0x24b6,0x24cf,1,26},
3274 	{0x2c00,0x2c2e,1,48},
3275 	{0x2c60,0x2c60,-1,1},
3276 	{0x2c62,0x2c62,-1,-10743},
3277 	{0x2c63,0x2c63,-1,-3814},
3278 	{0x2c64,0x2c64,-1,-10727},
3279 	{0x2c67,0x2c6b,2,1},
3280 	{0x2c6d,0x2c6d,-1,-10780},
3281 	{0x2c6e,0x2c6e,-1,-10749},
3282 	{0x2c6f,0x2c6f,-1,-10783},
3283 	{0x2c70,0x2c70,-1,-10782},
3284 	{0x2c72,0x2c75,3,1},
3285 	{0x2c7e,0x2c7f,1,-10815},
3286 	{0x2c80,0x2ce2,2,1},
3287 	{0x2ceb,0x2ced,2,1},
3288 	{0x2cf2,0xa640,31054,1},
3289 	{0xa642,0xa66c,2,1},
3290 	{0xa680,0xa69a,2,1},
3291 	{0xa722,0xa72e,2,1},
3292 	{0xa732,0xa76e,2,1},
3293 	{0xa779,0xa77b,2,1},
3294 	{0xa77d,0xa77d,-1,-35332},
3295 	{0xa77e,0xa786,2,1},
3296 	{0xa78b,0xa78b,-1,1},
3297 	{0xa78d,0xa78d,-1,-42280},
3298 	{0xa790,0xa792,2,1},
3299 	{0xa796,0xa7a8,2,1},
3300 	{0xa7aa,0xa7aa,-1,-42308},
3301 	{0xa7ab,0xa7ab,-1,-42319},
3302 	{0xa7ac,0xa7ac,-1,-42315},
3303 	{0xa7ad,0xa7ad,-1,-42305},
3304 	{0xa7ae,0xa7ae,-1,-42308},
3305 	{0xa7b0,0xa7b0,-1,-42258},
3306 	{0xa7b1,0xa7b1,-1,-42282},
3307 	{0xa7b2,0xa7b2,-1,-42261},
3308 	{0xa7b3,0xa7b3,-1,928},
3309 	{0xa7b4,0xa7b8,2,1},
3310 	{0xff21,0xff3a,1,32},
3311 	{0x10400,0x10427,1,40},
3312 	{0x104b0,0x104d3,1,40},
3313 	{0x10c80,0x10cb2,1,64},
3314 	{0x118a0,0x118bf,1,32},
3315 	{0x16e40,0x16e5f,1,32},
3316 	{0x1e900,0x1e921,1,34}
3317 };
3318 
3319 static convertStruct toUpper[] =
3320 {
3321 	{0x61,0x7a,1,-32},
3322 	{0xb5,0xb5,-1,743},
3323 	{0xe0,0xf6,1,-32},
3324 	{0xf8,0xfe,1,-32},
3325 	{0xff,0xff,-1,121},
3326 	{0x101,0x12f,2,-1},
3327 	{0x131,0x131,-1,-232},
3328 	{0x133,0x137,2,-1},
3329 	{0x13a,0x148,2,-1},
3330 	{0x14b,0x177,2,-1},
3331 	{0x17a,0x17e,2,-1},
3332 	{0x17f,0x17f,-1,-300},
3333 	{0x180,0x180,-1,195},
3334 	{0x183,0x185,2,-1},
3335 	{0x188,0x18c,4,-1},
3336 	{0x192,0x192,-1,-1},
3337 	{0x195,0x195,-1,97},
3338 	{0x199,0x199,-1,-1},
3339 	{0x19a,0x19a,-1,163},
3340 	{0x19e,0x19e,-1,130},
3341 	{0x1a1,0x1a5,2,-1},
3342 	{0x1a8,0x1ad,5,-1},
3343 	{0x1b0,0x1b4,4,-1},
3344 	{0x1b6,0x1b9,3,-1},
3345 	{0x1bd,0x1bd,-1,-1},
3346 	{0x1bf,0x1bf,-1,56},
3347 	{0x1c5,0x1c5,-1,-1},
3348 	{0x1c6,0x1c6,-1,-2},
3349 	{0x1c8,0x1c8,-1,-1},
3350 	{0x1c9,0x1c9,-1,-2},
3351 	{0x1cb,0x1cb,-1,-1},
3352 	{0x1cc,0x1cc,-1,-2},
3353 	{0x1ce,0x1dc,2,-1},
3354 	{0x1dd,0x1dd,-1,-79},
3355 	{0x1df,0x1ef,2,-1},
3356 	{0x1f2,0x1f2,-1,-1},
3357 	{0x1f3,0x1f3,-1,-2},
3358 	{0x1f5,0x1f9,4,-1},
3359 	{0x1fb,0x21f,2,-1},
3360 	{0x223,0x233,2,-1},
3361 	{0x23c,0x23c,-1,-1},
3362 	{0x23f,0x240,1,10815},
3363 	{0x242,0x247,5,-1},
3364 	{0x249,0x24f,2,-1},
3365 	{0x250,0x250,-1,10783},
3366 	{0x251,0x251,-1,10780},
3367 	{0x252,0x252,-1,10782},
3368 	{0x253,0x253,-1,-210},
3369 	{0x254,0x254,-1,-206},
3370 	{0x256,0x257,1,-205},
3371 	{0x259,0x259,-1,-202},
3372 	{0x25b,0x25b,-1,-203},
3373 	{0x25c,0x25c,-1,42319},
3374 	{0x260,0x260,-1,-205},
3375 	{0x261,0x261,-1,42315},
3376 	{0x263,0x263,-1,-207},
3377 	{0x265,0x265,-1,42280},
3378 	{0x266,0x266,-1,42308},
3379 	{0x268,0x268,-1,-209},
3380 	{0x269,0x269,-1,-211},
3381 	{0x26a,0x26a,-1,42308},
3382 	{0x26b,0x26b,-1,10743},
3383 	{0x26c,0x26c,-1,42305},
3384 	{0x26f,0x26f,-1,-211},
3385 	{0x271,0x271,-1,10749},
3386 	{0x272,0x272,-1,-213},
3387 	{0x275,0x275,-1,-214},
3388 	{0x27d,0x27d,-1,10727},
3389 	{0x280,0x283,3,-218},
3390 	{0x287,0x287,-1,42282},
3391 	{0x288,0x288,-1,-218},
3392 	{0x289,0x289,-1,-69},
3393 	{0x28a,0x28b,1,-217},
3394 	{0x28c,0x28c,-1,-71},
3395 	{0x292,0x292,-1,-219},
3396 	{0x29d,0x29d,-1,42261},
3397 	{0x29e,0x29e,-1,42258},
3398 	{0x345,0x345,-1,84},
3399 	{0x371,0x373,2,-1},
3400 	{0x377,0x377,-1,-1},
3401 	{0x37b,0x37d,1,130},
3402 	{0x3ac,0x3ac,-1,-38},
3403 	{0x3ad,0x3af,1,-37},
3404 	{0x3b1,0x3c1,1,-32},
3405 	{0x3c2,0x3c2,-1,-31},
3406 	{0x3c3,0x3cb,1,-32},
3407 	{0x3cc,0x3cc,-1,-64},
3408 	{0x3cd,0x3ce,1,-63},
3409 	{0x3d0,0x3d0,-1,-62},
3410 	{0x3d1,0x3d1,-1,-57},
3411 	{0x3d5,0x3d5,-1,-47},
3412 	{0x3d6,0x3d6,-1,-54},
3413 	{0x3d7,0x3d7,-1,-8},
3414 	{0x3d9,0x3ef,2,-1},
3415 	{0x3f0,0x3f0,-1,-86},
3416 	{0x3f1,0x3f1,-1,-80},
3417 	{0x3f2,0x3f2,-1,7},
3418 	{0x3f3,0x3f3,-1,-116},
3419 	{0x3f5,0x3f5,-1,-96},
3420 	{0x3f8,0x3fb,3,-1},
3421 	{0x430,0x44f,1,-32},
3422 	{0x450,0x45f,1,-80},
3423 	{0x461,0x481,2,-1},
3424 	{0x48b,0x4bf,2,-1},
3425 	{0x4c2,0x4ce,2,-1},
3426 	{0x4cf,0x4cf,-1,-15},
3427 	{0x4d1,0x52f,2,-1},
3428 	{0x561,0x586,1,-48},
3429 	{0x10d0,0x10fa,1,3008},
3430 	{0x10fd,0x10ff,1,3008},
3431 	{0x13f8,0x13fd,1,-8},
3432 	{0x1c80,0x1c80,-1,-6254},
3433 	{0x1c81,0x1c81,-1,-6253},
3434 	{0x1c82,0x1c82,-1,-6244},
3435 	{0x1c83,0x1c84,1,-6242},
3436 	{0x1c85,0x1c85,-1,-6243},
3437 	{0x1c86,0x1c86,-1,-6236},
3438 	{0x1c87,0x1c87,-1,-6181},
3439 	{0x1c88,0x1c88,-1,35266},
3440 	{0x1d79,0x1d79,-1,35332},
3441 	{0x1d7d,0x1d7d,-1,3814},
3442 	{0x1e01,0x1e95,2,-1},
3443 	{0x1e9b,0x1e9b,-1,-59},
3444 	{0x1ea1,0x1eff,2,-1},
3445 	{0x1f00,0x1f07,1,8},
3446 	{0x1f10,0x1f15,1,8},
3447 	{0x1f20,0x1f27,1,8},
3448 	{0x1f30,0x1f37,1,8},
3449 	{0x1f40,0x1f45,1,8},
3450 	{0x1f51,0x1f57,2,8},
3451 	{0x1f60,0x1f67,1,8},
3452 	{0x1f70,0x1f71,1,74},
3453 	{0x1f72,0x1f75,1,86},
3454 	{0x1f76,0x1f77,1,100},
3455 	{0x1f78,0x1f79,1,128},
3456 	{0x1f7a,0x1f7b,1,112},
3457 	{0x1f7c,0x1f7d,1,126},
3458 	{0x1f80,0x1f87,1,8},
3459 	{0x1f90,0x1f97,1,8},
3460 	{0x1fa0,0x1fa7,1,8},
3461 	{0x1fb0,0x1fb1,1,8},
3462 	{0x1fb3,0x1fb3,-1,9},
3463 	{0x1fbe,0x1fbe,-1,-7205},
3464 	{0x1fc3,0x1fc3,-1,9},
3465 	{0x1fd0,0x1fd1,1,8},
3466 	{0x1fe0,0x1fe1,1,8},
3467 	{0x1fe5,0x1fe5,-1,7},
3468 	{0x1ff3,0x1ff3,-1,9},
3469 	{0x214e,0x214e,-1,-28},
3470 	{0x2170,0x217f,1,-16},
3471 	{0x2184,0x2184,-1,-1},
3472 	{0x24d0,0x24e9,1,-26},
3473 	{0x2c30,0x2c5e,1,-48},
3474 	{0x2c61,0x2c61,-1,-1},
3475 	{0x2c65,0x2c65,-1,-10795},
3476 	{0x2c66,0x2c66,-1,-10792},
3477 	{0x2c68,0x2c6c,2,-1},
3478 	{0x2c73,0x2c76,3,-1},
3479 	{0x2c81,0x2ce3,2,-1},
3480 	{0x2cec,0x2cee,2,-1},
3481 	{0x2cf3,0x2cf3,-1,-1},
3482 	{0x2d00,0x2d25,1,-7264},
3483 	{0x2d27,0x2d2d,6,-7264},
3484 	{0xa641,0xa66d,2,-1},
3485 	{0xa681,0xa69b,2,-1},
3486 	{0xa723,0xa72f,2,-1},
3487 	{0xa733,0xa76f,2,-1},
3488 	{0xa77a,0xa77c,2,-1},
3489 	{0xa77f,0xa787,2,-1},
3490 	{0xa78c,0xa791,5,-1},
3491 	{0xa793,0xa797,4,-1},
3492 	{0xa799,0xa7a9,2,-1},
3493 	{0xa7b5,0xa7b9,2,-1},
3494 	{0xab53,0xab53,-1,-928},
3495 	{0xab70,0xabbf,1,-38864},
3496 	{0xff41,0xff5a,1,-32},
3497 	{0x10428,0x1044f,1,-40},
3498 	{0x104d8,0x104fb,1,-40},
3499 	{0x10cc0,0x10cf2,1,-64},
3500 	{0x118c0,0x118df,1,-32},
3501 	{0x16e60,0x16e7f,1,-32},
3502 	{0x1e922,0x1e943,1,-34}
3503 };
3504 
3505 /*
3506  * Return the upper-case equivalent of "a", which is a UCS-4 character.  Use
3507  * simple case folding.
3508  */
3509     int
3510 utf_toupper(int a)
3511 {
3512     /* If 'casemap' contains "keepascii" use ASCII style toupper(). */
3513     if (a < 128 && (cmp_flags & CMP_KEEPASCII))
3514 	return TOUPPER_ASC(a);
3515 
3516 #if defined(HAVE_TOWUPPER) && defined(__STDC_ISO_10646__)
3517     /* If towupper() is available and handles Unicode, use it. */
3518     if (!(cmp_flags & CMP_INTERNAL))
3519 	return towupper(a);
3520 #endif
3521 
3522     /* For characters below 128 use locale sensitive toupper(). */
3523     if (a < 128)
3524 	return TOUPPER_LOC(a);
3525 
3526     /* For any other characters use the above mapping table. */
3527     return utf_convert(a, toUpper, (int)sizeof(toUpper));
3528 }
3529 
3530     int
3531 utf_islower(int a)
3532 {
3533     /* German sharp s is lower case but has no upper case equivalent. */
3534     return (utf_toupper(a) != a) || a == 0xdf;
3535 }
3536 
3537 /*
3538  * Return the lower-case equivalent of "a", which is a UCS-4 character.  Use
3539  * simple case folding.
3540  */
3541     int
3542 utf_tolower(int a)
3543 {
3544     /* If 'casemap' contains "keepascii" use ASCII style tolower(). */
3545     if (a < 128 && (cmp_flags & CMP_KEEPASCII))
3546 	return TOLOWER_ASC(a);
3547 
3548 #if defined(HAVE_TOWLOWER) && defined(__STDC_ISO_10646__)
3549     /* If towlower() is available and handles Unicode, use it. */
3550     if (!(cmp_flags & CMP_INTERNAL))
3551 	return towlower(a);
3552 #endif
3553 
3554     /* For characters below 128 use locale sensitive tolower(). */
3555     if (a < 128)
3556 	return TOLOWER_LOC(a);
3557 
3558     /* For any other characters use the above mapping table. */
3559     return utf_convert(a, toLower, (int)sizeof(toLower));
3560 }
3561 
3562     int
3563 utf_isupper(int a)
3564 {
3565     return (utf_tolower(a) != a);
3566 }
3567 
3568     static int
3569 utf_strnicmp(
3570     char_u      *s1,
3571     char_u      *s2,
3572     size_t      n1,
3573     size_t      n2)
3574 {
3575     int		c1, c2, cdiff;
3576     char_u	buffer[6];
3577 
3578     for (;;)
3579     {
3580 	c1 = utf_safe_read_char_adv(&s1, &n1);
3581 	c2 = utf_safe_read_char_adv(&s2, &n2);
3582 
3583 	if (c1 <= 0 || c2 <= 0)
3584 	    break;
3585 
3586 	if (c1 == c2)
3587 	    continue;
3588 
3589 	cdiff = utf_fold(c1) - utf_fold(c2);
3590 	if (cdiff != 0)
3591 	    return cdiff;
3592     }
3593 
3594     /* some string ended or has an incomplete/illegal character sequence */
3595 
3596     if (c1 == 0 || c2 == 0)
3597     {
3598 	/* some string ended. shorter string is smaller */
3599 	if (c1 == 0 && c2 == 0)
3600 	    return 0;
3601 	return c1 == 0 ? -1 : 1;
3602     }
3603 
3604     /* Continue with bytewise comparison to produce some result that
3605      * would make comparison operations involving this function transitive.
3606      *
3607      * If only one string had an error, comparison should be made with
3608      * folded version of the other string. In this case it is enough
3609      * to fold just one character to determine the result of comparison. */
3610 
3611     if (c1 != -1 && c2 == -1)
3612     {
3613 	n1 = utf_char2bytes(utf_fold(c1), buffer);
3614 	s1 = buffer;
3615     }
3616     else if (c2 != -1 && c1 == -1)
3617     {
3618 	n2 = utf_char2bytes(utf_fold(c2), buffer);
3619 	s2 = buffer;
3620     }
3621 
3622     while (n1 > 0 && n2 > 0 && *s1 != NUL && *s2 != NUL)
3623     {
3624 	cdiff = (int)(*s1) - (int)(*s2);
3625 	if (cdiff != 0)
3626 	    return cdiff;
3627 
3628 	s1++;
3629 	s2++;
3630 	n1--;
3631 	n2--;
3632     }
3633 
3634     if (n1 > 0 && *s1 == NUL)
3635 	n1 = 0;
3636     if (n2 > 0 && *s2 == NUL)
3637 	n2 = 0;
3638 
3639     if (n1 == 0 && n2 == 0)
3640 	return 0;
3641     return n1 == 0 ? -1 : 1;
3642 }
3643 
3644 /*
3645  * Version of strnicmp() that handles multi-byte characters.
3646  * Needed for Big5, Shift-JIS and UTF-8 encoding.  Other DBCS encodings can
3647  * probably use strnicmp(), because there are no ASCII characters in the
3648  * second byte.
3649  * Returns zero if s1 and s2 are equal (ignoring case), the difference between
3650  * two characters otherwise.
3651  */
3652     int
3653 mb_strnicmp(char_u *s1, char_u *s2, size_t nn)
3654 {
3655     int		i, l;
3656     int		cdiff;
3657     int		n = (int)nn;
3658 
3659     if (enc_utf8)
3660     {
3661 	return utf_strnicmp(s1, s2, nn, nn);
3662     }
3663     else
3664     {
3665 	for (i = 0; i < n; i += l)
3666 	{
3667 	    if (s1[i] == NUL && s2[i] == NUL)	/* both strings end */
3668 		return 0;
3669 
3670 	    l = (*mb_ptr2len)(s1 + i);
3671 	    if (l <= 1)
3672 	    {
3673 		/* Single byte: first check normally, then with ignore case. */
3674 		if (s1[i] != s2[i])
3675 		{
3676 		    cdiff = MB_TOLOWER(s1[i]) - MB_TOLOWER(s2[i]);
3677 		    if (cdiff != 0)
3678 			return cdiff;
3679 		}
3680 	    }
3681 	    else
3682 	    {
3683 		/* For non-Unicode multi-byte don't ignore case. */
3684 		if (l > n - i)
3685 		    l = n - i;
3686 		cdiff = STRNCMP(s1 + i, s2 + i, l);
3687 		if (cdiff != 0)
3688 		    return cdiff;
3689 	    }
3690 	}
3691     }
3692     return 0;
3693 }
3694 
3695 /*
3696  * "g8": show bytes of the UTF-8 char under the cursor.  Doesn't matter what
3697  * 'encoding' has been set to.
3698  */
3699     void
3700 show_utf8(void)
3701 {
3702     int		len;
3703     int		rlen = 0;
3704     char_u	*line;
3705     int		clen;
3706     int		i;
3707 
3708     /* Get the byte length of the char under the cursor, including composing
3709      * characters. */
3710     line = ml_get_cursor();
3711     len = utfc_ptr2len(line);
3712     if (len == 0)
3713     {
3714 	msg("NUL");
3715 	return;
3716     }
3717 
3718     clen = 0;
3719     for (i = 0; i < len; ++i)
3720     {
3721 	if (clen == 0)
3722 	{
3723 	    /* start of (composing) character, get its length */
3724 	    if (i > 0)
3725 	    {
3726 		STRCPY(IObuff + rlen, "+ ");
3727 		rlen += 2;
3728 	    }
3729 	    clen = utf_ptr2len(line + i);
3730 	}
3731 	sprintf((char *)IObuff + rlen, "%02x ",
3732 		(line[i] == NL) ? NUL : line[i]);  /* NUL is stored as NL */
3733 	--clen;
3734 	rlen += (int)STRLEN(IObuff + rlen);
3735 	if (rlen > IOSIZE - 20)
3736 	    break;
3737     }
3738 
3739     msg((char *)IObuff);
3740 }
3741 
3742 /*
3743  * mb_head_off() function pointer.
3744  * Return offset from "p" to the first byte of the character it points into.
3745  * If "p" points to the NUL at the end of the string return 0.
3746  * Returns 0 when already at the first byte of a character.
3747  */
3748     int
3749 latin_head_off(char_u *base UNUSED, char_u *p UNUSED)
3750 {
3751     return 0;
3752 }
3753 
3754     int
3755 dbcs_head_off(char_u *base, char_u *p)
3756 {
3757     char_u	*q;
3758 
3759     /* It can't be a trailing byte when not using DBCS, at the start of the
3760      * string or the previous byte can't start a double-byte. */
3761     if (p <= base || MB_BYTE2LEN(p[-1]) == 1 || *p == NUL)
3762 	return 0;
3763 
3764     /* This is slow: need to start at the base and go forward until the
3765      * byte we are looking for.  Return 1 when we went past it, 0 otherwise. */
3766     q = base;
3767     while (q < p)
3768 	q += dbcs_ptr2len(q);
3769     return (q == p) ? 0 : 1;
3770 }
3771 
3772 /*
3773  * Special version of dbcs_head_off() that works for ScreenLines[], where
3774  * single-width DBCS_JPNU characters are stored separately.
3775  */
3776     int
3777 dbcs_screen_head_off(char_u *base, char_u *p)
3778 {
3779     char_u	*q;
3780 
3781     /* It can't be a trailing byte when not using DBCS, at the start of the
3782      * string or the previous byte can't start a double-byte.
3783      * For euc-jp an 0x8e byte in the previous cell always means we have a
3784      * lead byte in the current cell. */
3785     if (p <= base
3786 	    || (enc_dbcs == DBCS_JPNU && p[-1] == 0x8e)
3787 	    || MB_BYTE2LEN(p[-1]) == 1
3788 	    || *p == NUL)
3789 	return 0;
3790 
3791     /* This is slow: need to start at the base and go forward until the
3792      * byte we are looking for.  Return 1 when we went past it, 0 otherwise.
3793      * For DBCS_JPNU look out for 0x8e, which means the second byte is not
3794      * stored as the next byte. */
3795     q = base;
3796     while (q < p)
3797     {
3798 	if (enc_dbcs == DBCS_JPNU && *q == 0x8e)
3799 	    ++q;
3800 	else
3801 	    q += dbcs_ptr2len(q);
3802     }
3803     return (q == p) ? 0 : 1;
3804 }
3805 
3806     int
3807 utf_head_off(char_u *base, char_u *p)
3808 {
3809     char_u	*q;
3810     char_u	*s;
3811     int		c;
3812     int		len;
3813 #ifdef FEAT_ARABIC
3814     char_u	*j;
3815 #endif
3816 
3817     if (*p < 0x80)		/* be quick for ASCII */
3818 	return 0;
3819 
3820     /* Skip backwards over trailing bytes: 10xx.xxxx
3821      * Skip backwards again if on a composing char. */
3822     for (q = p; ; --q)
3823     {
3824 	/* Move s to the last byte of this char. */
3825 	for (s = q; (s[1] & 0xc0) == 0x80; ++s)
3826 	    ;
3827 	/* Move q to the first byte of this char. */
3828 	while (q > base && (*q & 0xc0) == 0x80)
3829 	    --q;
3830 	/* Check for illegal sequence. Do allow an illegal byte after where we
3831 	 * started. */
3832 	len = utf8len_tab[*q];
3833 	if (len != (int)(s - q + 1) && len != (int)(p - q + 1))
3834 	    return 0;
3835 
3836 	if (q <= base)
3837 	    break;
3838 
3839 	c = utf_ptr2char(q);
3840 	if (utf_iscomposing(c))
3841 	    continue;
3842 
3843 #ifdef FEAT_ARABIC
3844 	if (arabic_maycombine(c))
3845 	{
3846 	    /* Advance to get a sneak-peak at the next char */
3847 	    j = q;
3848 	    --j;
3849 	    /* Move j to the first byte of this char. */
3850 	    while (j > base && (*j & 0xc0) == 0x80)
3851 		--j;
3852 	    if (arabic_combine(utf_ptr2char(j), c))
3853 		continue;
3854 	}
3855 #endif
3856 	break;
3857     }
3858 
3859     return (int)(p - q);
3860 }
3861 
3862 /*
3863  * Copy a character from "*fp" to "*tp" and advance the pointers.
3864  */
3865     void
3866 mb_copy_char(char_u **fp, char_u **tp)
3867 {
3868     int	    l = (*mb_ptr2len)(*fp);
3869 
3870     mch_memmove(*tp, *fp, (size_t)l);
3871     *tp += l;
3872     *fp += l;
3873 }
3874 
3875 /*
3876  * Return the offset from "p" to the first byte of a character.  When "p" is
3877  * at the start of a character 0 is returned, otherwise the offset to the next
3878  * character.  Can start anywhere in a stream of bytes.
3879  */
3880     int
3881 mb_off_next(char_u *base, char_u *p)
3882 {
3883     int		i;
3884     int		j;
3885 
3886     if (enc_utf8)
3887     {
3888 	if (*p < 0x80)		/* be quick for ASCII */
3889 	    return 0;
3890 
3891 	/* Find the next character that isn't 10xx.xxxx */
3892 	for (i = 0; (p[i] & 0xc0) == 0x80; ++i)
3893 	    ;
3894 	if (i > 0)
3895 	{
3896 	    /* Check for illegal sequence. */
3897 	    for (j = 0; p - j > base; ++j)
3898 		if ((p[-j] & 0xc0) != 0x80)
3899 		    break;
3900 	    if (utf8len_tab[p[-j]] != i + j)
3901 		return 0;
3902 	}
3903 	return i;
3904     }
3905 
3906     /* Only need to check if we're on a trail byte, it doesn't matter if we
3907      * want the offset to the next or current character. */
3908     return (*mb_head_off)(base, p);
3909 }
3910 
3911 /*
3912  * Return the offset from "p" to the last byte of the character it points
3913  * into.  Can start anywhere in a stream of bytes.
3914  */
3915     int
3916 mb_tail_off(char_u *base, char_u *p)
3917 {
3918     int		i;
3919     int		j;
3920 
3921     if (*p == NUL)
3922 	return 0;
3923 
3924     if (enc_utf8)
3925     {
3926 	/* Find the last character that is 10xx.xxxx */
3927 	for (i = 0; (p[i + 1] & 0xc0) == 0x80; ++i)
3928 	    ;
3929 	/* Check for illegal sequence. */
3930 	for (j = 0; p - j > base; ++j)
3931 	    if ((p[-j] & 0xc0) != 0x80)
3932 		break;
3933 	if (utf8len_tab[p[-j]] != i + j + 1)
3934 	    return 0;
3935 	return i;
3936     }
3937 
3938     /* It can't be the first byte if a double-byte when not using DBCS, at the
3939      * end of the string or the byte can't start a double-byte. */
3940     if (enc_dbcs == 0 || p[1] == NUL || MB_BYTE2LEN(*p) == 1)
3941 	return 0;
3942 
3943     /* Return 1 when on the lead byte, 0 when on the tail byte. */
3944     return 1 - dbcs_head_off(base, p);
3945 }
3946 
3947 /*
3948  * Find the next illegal byte sequence.
3949  */
3950     void
3951 utf_find_illegal(void)
3952 {
3953     pos_T	pos = curwin->w_cursor;
3954     char_u	*p;
3955     int		len;
3956     vimconv_T	vimconv;
3957     char_u	*tofree = NULL;
3958 
3959     vimconv.vc_type = CONV_NONE;
3960     if (enc_utf8 && (enc_canon_props(curbuf->b_p_fenc) & ENC_8BIT))
3961     {
3962 	/* 'encoding' is "utf-8" but we are editing a 8-bit encoded file,
3963 	 * possibly a utf-8 file with illegal bytes.  Setup for conversion
3964 	 * from utf-8 to 'fileencoding'. */
3965 	convert_setup(&vimconv, p_enc, curbuf->b_p_fenc);
3966     }
3967 
3968     curwin->w_cursor.coladd = 0;
3969     for (;;)
3970     {
3971 	p = ml_get_cursor();
3972 	if (vimconv.vc_type != CONV_NONE)
3973 	{
3974 	    vim_free(tofree);
3975 	    tofree = string_convert(&vimconv, p, NULL);
3976 	    if (tofree == NULL)
3977 		break;
3978 	    p = tofree;
3979 	}
3980 
3981 	while (*p != NUL)
3982 	{
3983 	    /* Illegal means that there are not enough trail bytes (checked by
3984 	     * utf_ptr2len()) or too many of them (overlong sequence). */
3985 	    len = utf_ptr2len(p);
3986 	    if (*p >= 0x80 && (len == 1
3987 				     || utf_char2len(utf_ptr2char(p)) != len))
3988 	    {
3989 		if (vimconv.vc_type == CONV_NONE)
3990 		    curwin->w_cursor.col += (colnr_T)(p - ml_get_cursor());
3991 		else
3992 		{
3993 		    int	    l;
3994 
3995 		    len = (int)(p - tofree);
3996 		    for (p = ml_get_cursor(); *p != NUL && len-- > 0; p += l)
3997 		    {
3998 			l = utf_ptr2len(p);
3999 			curwin->w_cursor.col += l;
4000 		    }
4001 		}
4002 		goto theend;
4003 	    }
4004 	    p += len;
4005 	}
4006 	if (curwin->w_cursor.lnum == curbuf->b_ml.ml_line_count)
4007 	    break;
4008 	++curwin->w_cursor.lnum;
4009 	curwin->w_cursor.col = 0;
4010     }
4011 
4012     /* didn't find it: don't move and beep */
4013     curwin->w_cursor = pos;
4014     beep_flush();
4015 
4016 theend:
4017     vim_free(tofree);
4018     convert_setup(&vimconv, NULL, NULL);
4019 }
4020 
4021 #if defined(FEAT_GUI_GTK) || defined(PROTO)
4022 /*
4023  * Return TRUE if string "s" is a valid utf-8 string.
4024  * When "end" is NULL stop at the first NUL.
4025  * When "end" is positive stop there.
4026  */
4027     int
4028 utf_valid_string(char_u *s, char_u *end)
4029 {
4030     int		l;
4031     char_u	*p = s;
4032 
4033     while (end == NULL ? *p != NUL : p < end)
4034     {
4035 	l = utf8len_tab_zero[*p];
4036 	if (l == 0)
4037 	    return FALSE;	/* invalid lead byte */
4038 	if (end != NULL && p + l > end)
4039 	    return FALSE;	/* incomplete byte sequence */
4040 	++p;
4041 	while (--l > 0)
4042 	    if ((*p++ & 0xc0) != 0x80)
4043 		return FALSE;	/* invalid trail byte */
4044     }
4045     return TRUE;
4046 }
4047 #endif
4048 
4049 #if defined(FEAT_GUI) || defined(PROTO)
4050 /*
4051  * Special version of mb_tail_off() for use in ScreenLines[].
4052  */
4053     int
4054 dbcs_screen_tail_off(char_u *base, char_u *p)
4055 {
4056     /* It can't be the first byte if a double-byte when not using DBCS, at the
4057      * end of the string or the byte can't start a double-byte.
4058      * For euc-jp an 0x8e byte always means we have a lead byte in the current
4059      * cell. */
4060     if (*p == NUL || p[1] == NUL
4061 	    || (enc_dbcs == DBCS_JPNU && *p == 0x8e)
4062 	    || MB_BYTE2LEN(*p) == 1)
4063 	return 0;
4064 
4065     /* Return 1 when on the lead byte, 0 when on the tail byte. */
4066     return 1 - dbcs_screen_head_off(base, p);
4067 }
4068 #endif
4069 
4070 /*
4071  * If the cursor moves on an trail byte, set the cursor on the lead byte.
4072  * Thus it moves left if necessary.
4073  * Return TRUE when the cursor was adjusted.
4074  */
4075     void
4076 mb_adjust_cursor(void)
4077 {
4078     mb_adjustpos(curbuf, &curwin->w_cursor);
4079 }
4080 
4081 /*
4082  * Adjust position "*lp" to point to the first byte of a multi-byte character.
4083  * If it points to a tail byte it's moved backwards to the head byte.
4084  */
4085     void
4086 mb_adjustpos(buf_T *buf, pos_T *lp)
4087 {
4088     char_u	*p;
4089 
4090     if (lp->col > 0 || lp->coladd > 1)
4091     {
4092 	p = ml_get_buf(buf, lp->lnum, FALSE);
4093 	if (*p == NUL || (int)STRLEN(p) < lp->col)
4094 	    lp->col = 0;
4095 	else
4096 	    lp->col -= (*mb_head_off)(p, p + lp->col);
4097 	/* Reset "coladd" when the cursor would be on the right half of a
4098 	 * double-wide character. */
4099 	if (lp->coladd == 1
4100 		&& p[lp->col] != TAB
4101 		&& vim_isprintc((*mb_ptr2char)(p + lp->col))
4102 		&& ptr2cells(p + lp->col) > 1)
4103 	    lp->coladd = 0;
4104     }
4105 }
4106 
4107 /*
4108  * Return a pointer to the character before "*p", if there is one.
4109  */
4110     char_u *
4111 mb_prevptr(
4112     char_u *line,	/* start of the string */
4113     char_u *p)
4114 {
4115     if (p > line)
4116 	MB_PTR_BACK(line, p);
4117     return p;
4118 }
4119 
4120 /*
4121  * Return the character length of "str".  Each multi-byte character (with
4122  * following composing characters) counts as one.
4123  */
4124     int
4125 mb_charlen(char_u *str)
4126 {
4127     char_u	*p = str;
4128     int		count;
4129 
4130     if (p == NULL)
4131 	return 0;
4132 
4133     for (count = 0; *p != NUL; count++)
4134 	p += (*mb_ptr2len)(p);
4135 
4136     return count;
4137 }
4138 
4139 #if defined(FEAT_SPELL) || defined(PROTO)
4140 /*
4141  * Like mb_charlen() but for a string with specified length.
4142  */
4143     int
4144 mb_charlen_len(char_u *str, int len)
4145 {
4146     char_u	*p = str;
4147     int		count;
4148 
4149     for (count = 0; *p != NUL && p < str + len; count++)
4150 	p += (*mb_ptr2len)(p);
4151 
4152     return count;
4153 }
4154 #endif
4155 
4156 /*
4157  * Try to un-escape a multi-byte character.
4158  * Used for the "to" and "from" part of a mapping.
4159  * Return the un-escaped string if it is a multi-byte character, and advance
4160  * "pp" to just after the bytes that formed it.
4161  * Return NULL if no multi-byte char was found.
4162  */
4163     char_u *
4164 mb_unescape(char_u **pp)
4165 {
4166     static char_u	buf[6];
4167     int			n;
4168     int			m = 0;
4169     char_u		*str = *pp;
4170 
4171     /* Must translate K_SPECIAL KS_SPECIAL KE_FILLER to K_SPECIAL and CSI
4172      * KS_EXTRA KE_CSI to CSI.
4173      * Maximum length of a utf-8 character is 4 bytes. */
4174     for (n = 0; str[n] != NUL && m < 4; ++n)
4175     {
4176 	if (str[n] == K_SPECIAL
4177 		&& str[n + 1] == KS_SPECIAL
4178 		&& str[n + 2] == KE_FILLER)
4179 	{
4180 	    buf[m++] = K_SPECIAL;
4181 	    n += 2;
4182 	}
4183 	else if ((str[n] == K_SPECIAL
4184 # ifdef FEAT_GUI
4185 		    || str[n] == CSI
4186 # endif
4187 		 )
4188 		&& str[n + 1] == KS_EXTRA
4189 		&& str[n + 2] == (int)KE_CSI)
4190 	{
4191 	    buf[m++] = CSI;
4192 	    n += 2;
4193 	}
4194 	else if (str[n] == K_SPECIAL
4195 # ifdef FEAT_GUI
4196 		|| str[n] == CSI
4197 # endif
4198 		)
4199 	    break;		/* a special key can't be a multibyte char */
4200 	else
4201 	    buf[m++] = str[n];
4202 	buf[m] = NUL;
4203 
4204 	/* Return a multi-byte character if it's found.  An illegal sequence
4205 	 * will result in a 1 here. */
4206 	if ((*mb_ptr2len)(buf) > 1)
4207 	{
4208 	    *pp = str + n + 1;
4209 	    return buf;
4210 	}
4211 
4212 	/* Bail out quickly for ASCII. */
4213 	if (buf[0] < 128)
4214 	    break;
4215     }
4216     return NULL;
4217 }
4218 
4219 /*
4220  * Return TRUE if the character at "row"/"col" on the screen is the left side
4221  * of a double-width character.
4222  * Caller must make sure "row" and "col" are not invalid!
4223  */
4224     int
4225 mb_lefthalve(int row, int col)
4226 {
4227 #ifdef FEAT_HANGULIN
4228     if (composing_hangul)
4229 	return TRUE;
4230 #endif
4231     return (*mb_off2cells)(LineOffset[row] + col,
4232 					LineOffset[row] + screen_Columns) > 1;
4233 }
4234 
4235 /*
4236  * Correct a position on the screen, if it's the right half of a double-wide
4237  * char move it to the left half.  Returns the corrected column.
4238  */
4239     int
4240 mb_fix_col(int col, int row)
4241 {
4242     col = check_col(col);
4243     row = check_row(row);
4244     if (has_mbyte && ScreenLines != NULL && col > 0
4245 	    && ((enc_dbcs
4246 		    && ScreenLines[LineOffset[row] + col] != NUL
4247 		    && dbcs_screen_head_off(ScreenLines + LineOffset[row],
4248 					 ScreenLines + LineOffset[row] + col))
4249 		|| (enc_utf8 && ScreenLines[LineOffset[row] + col] == 0)))
4250 	return col - 1;
4251     return col;
4252 }
4253 
4254 static int enc_alias_search(char_u *name);
4255 
4256 /*
4257  * Skip the Vim specific head of a 'encoding' name.
4258  */
4259     char_u *
4260 enc_skip(char_u *p)
4261 {
4262     if (STRNCMP(p, "2byte-", 6) == 0)
4263 	return p + 6;
4264     if (STRNCMP(p, "8bit-", 5) == 0)
4265 	return p + 5;
4266     return p;
4267 }
4268 
4269 /*
4270  * Find the canonical name for encoding "enc".
4271  * When the name isn't recognized, returns "enc" itself, but with all lower
4272  * case characters and '_' replaced with '-'.
4273  * Returns an allocated string.  NULL for out-of-memory.
4274  */
4275     char_u *
4276 enc_canonize(char_u *enc)
4277 {
4278     char_u	*r;
4279     char_u	*p, *s;
4280     int		i;
4281 
4282     if (STRCMP(enc, "default") == 0)
4283     {
4284 	/* Use the default encoding as it's found by set_init_1(). */
4285 	r = get_encoding_default();
4286 	if (r == NULL)
4287 	    r = (char_u *)"latin1";
4288 	return vim_strsave(r);
4289     }
4290 
4291     /* copy "enc" to allocated memory, with room for two '-' */
4292     r = alloc((unsigned)(STRLEN(enc) + 3));
4293     if (r != NULL)
4294     {
4295 	/* Make it all lower case and replace '_' with '-'. */
4296 	p = r;
4297 	for (s = enc; *s != NUL; ++s)
4298 	{
4299 	    if (*s == '_')
4300 		*p++ = '-';
4301 	    else
4302 		*p++ = TOLOWER_ASC(*s);
4303 	}
4304 	*p = NUL;
4305 
4306 	/* Skip "2byte-" and "8bit-". */
4307 	p = enc_skip(r);
4308 
4309 	/* Change "microsoft-cp" to "cp".  Used in some spell files. */
4310 	if (STRNCMP(p, "microsoft-cp", 12) == 0)
4311 	    STRMOVE(p, p + 10);
4312 
4313 	/* "iso8859" -> "iso-8859" */
4314 	if (STRNCMP(p, "iso8859", 7) == 0)
4315 	{
4316 	    STRMOVE(p + 4, p + 3);
4317 	    p[3] = '-';
4318 	}
4319 
4320 	/* "iso-8859n" -> "iso-8859-n" */
4321 	if (STRNCMP(p, "iso-8859", 8) == 0 && p[8] != '-')
4322 	{
4323 	    STRMOVE(p + 9, p + 8);
4324 	    p[8] = '-';
4325 	}
4326 
4327 	/* "latin-N" -> "latinN" */
4328 	if (STRNCMP(p, "latin-", 6) == 0)
4329 	    STRMOVE(p + 5, p + 6);
4330 
4331 	if (enc_canon_search(p) >= 0)
4332 	{
4333 	    /* canonical name can be used unmodified */
4334 	    if (p != r)
4335 		STRMOVE(r, p);
4336 	}
4337 	else if ((i = enc_alias_search(p)) >= 0)
4338 	{
4339 	    /* alias recognized, get canonical name */
4340 	    vim_free(r);
4341 	    r = vim_strsave((char_u *)enc_canon_table[i].name);
4342 	}
4343     }
4344     return r;
4345 }
4346 
4347 /*
4348  * Search for an encoding alias of "name".
4349  * Returns -1 when not found.
4350  */
4351     static int
4352 enc_alias_search(char_u *name)
4353 {
4354     int		i;
4355 
4356     for (i = 0; enc_alias_table[i].name != NULL; ++i)
4357 	if (STRCMP(name, enc_alias_table[i].name) == 0)
4358 	    return enc_alias_table[i].canon;
4359     return -1;
4360 }
4361 
4362 
4363 #ifdef HAVE_LANGINFO_H
4364 # include <langinfo.h>
4365 #endif
4366 
4367 #ifndef FEAT_GUI_MSWIN
4368 /*
4369  * Get the canonicalized encoding from the specified locale string "locale"
4370  * or from the environment variables LC_ALL, LC_CTYPE and LANG.
4371  * Returns an allocated string when successful, NULL when not.
4372  */
4373     char_u *
4374 enc_locale_env(char *locale)
4375 {
4376     char	*s = locale;
4377     char	*p;
4378     int		i;
4379     char	buf[50];
4380 
4381     if (s == NULL || *s == NUL)
4382 	if ((s = getenv("LC_ALL")) == NULL || *s == NUL)
4383 	    if ((s = getenv("LC_CTYPE")) == NULL || *s == NUL)
4384 		s = getenv("LANG");
4385 
4386     if (s == NULL || *s == NUL)
4387 	return NULL;
4388 
4389     /* The most generic locale format is:
4390      * language[_territory][.codeset][@modifier][+special][,[sponsor][_revision]]
4391      * If there is a '.' remove the part before it.
4392      * if there is something after the codeset, remove it.
4393      * Make the name lowercase and replace '_' with '-'.
4394      * Exception: "ja_JP.EUC" == "euc-jp", "zh_CN.EUC" = "euc-cn",
4395      * "ko_KR.EUC" == "euc-kr"
4396      */
4397     if ((p = (char *)vim_strchr((char_u *)s, '.')) != NULL)
4398     {
4399 	if (p > s + 2 && STRNICMP(p + 1, "EUC", 3) == 0
4400 			&& !isalnum((int)p[4]) && p[4] != '-' && p[-3] == '_')
4401 	{
4402 	    /* copy "XY.EUC" to "euc-XY" to buf[10] */
4403 	    STRCPY(buf + 10, "euc-");
4404 	    buf[14] = p[-2];
4405 	    buf[15] = p[-1];
4406 	    buf[16] = 0;
4407 	    s = buf + 10;
4408 	}
4409 	else
4410 	    s = p + 1;
4411     }
4412     for (i = 0; i < (int)sizeof(buf) - 1 && s[i] != NUL; ++i)
4413     {
4414 	if (s[i] == '_' || s[i] == '-')
4415 	    buf[i] = '-';
4416 	else if (isalnum((int)s[i]))
4417 	    buf[i] = TOLOWER_ASC(s[i]);
4418 	else
4419 	    break;
4420     }
4421     buf[i] = NUL;
4422 
4423     return enc_canonize((char_u *)buf);
4424 }
4425 #endif
4426 
4427 /*
4428  * Get the canonicalized encoding of the current locale.
4429  * Returns an allocated string when successful, NULL when not.
4430  */
4431     char_u *
4432 enc_locale(void)
4433 {
4434 #ifdef MSWIN
4435     char	buf[50];
4436     long	acp = GetACP();
4437 
4438     if (acp == 1200)
4439 	STRCPY(buf, "ucs-2le");
4440     else if (acp == 1252)	    /* cp1252 is used as latin1 */
4441 	STRCPY(buf, "latin1");
4442     else
4443 	sprintf(buf, "cp%ld", acp);
4444 
4445     return enc_canonize((char_u *)buf);
4446 #else
4447     char	*s;
4448 
4449 # ifdef HAVE_NL_LANGINFO_CODESET
4450     if ((s = nl_langinfo(CODESET)) == NULL || *s == NUL)
4451 # endif
4452 # if defined(HAVE_LOCALE_H) || defined(X_LOCALE)
4453 	if ((s = setlocale(LC_CTYPE, NULL)) == NULL || *s == NUL)
4454 # endif
4455 	    s = NULL;
4456 
4457     return enc_locale_env(s);
4458 #endif
4459 }
4460 
4461 # if defined(MSWIN) || defined(PROTO) || defined(FEAT_CYGWIN_WIN32_CLIPBOARD)
4462 /*
4463  * Convert an encoding name to an MS-Windows codepage.
4464  * Returns zero if no codepage can be figured out.
4465  */
4466     int
4467 encname2codepage(char_u *name)
4468 {
4469     int		cp;
4470     char_u	*p = name;
4471     int		idx;
4472 
4473     if (STRNCMP(p, "8bit-", 5) == 0)
4474 	p += 5;
4475     else if (STRNCMP(p_enc, "2byte-", 6) == 0)
4476 	p += 6;
4477 
4478     if (p[0] == 'c' && p[1] == 'p')
4479 	cp = atoi((char *)p + 2);
4480     else if ((idx = enc_canon_search(p)) >= 0)
4481 	cp = enc_canon_table[idx].codepage;
4482     else
4483 	return 0;
4484     if (IsValidCodePage(cp))
4485 	return cp;
4486     return 0;
4487 }
4488 # endif
4489 
4490 # if defined(USE_ICONV) || defined(PROTO)
4491 
4492 /*
4493  * Call iconv_open() with a check if iconv() works properly (there are broken
4494  * versions).
4495  * Returns (void *)-1 if failed.
4496  * (should return iconv_t, but that causes problems with prototypes).
4497  */
4498     void *
4499 my_iconv_open(char_u *to, char_u *from)
4500 {
4501     iconv_t	fd;
4502 #define ICONV_TESTLEN 400
4503     char_u	tobuf[ICONV_TESTLEN];
4504     char	*p;
4505     size_t	tolen;
4506     static int	iconv_ok = -1;
4507 
4508     if (iconv_ok == FALSE)
4509 	return (void *)-1;	/* detected a broken iconv() previously */
4510 
4511 #ifdef DYNAMIC_ICONV
4512     /* Check if the iconv.dll can be found. */
4513     if (!iconv_enabled(TRUE))
4514 	return (void *)-1;
4515 #endif
4516 
4517     fd = iconv_open((char *)enc_skip(to), (char *)enc_skip(from));
4518 
4519     if (fd != (iconv_t)-1 && iconv_ok == -1)
4520     {
4521 	/*
4522 	 * Do a dummy iconv() call to check if it actually works.  There is a
4523 	 * version of iconv() on Linux that is broken.  We can't ignore it,
4524 	 * because it's wide-spread.  The symptoms are that after outputting
4525 	 * the initial shift state the "to" pointer is NULL and conversion
4526 	 * stops for no apparent reason after about 8160 characters.
4527 	 */
4528 	p = (char *)tobuf;
4529 	tolen = ICONV_TESTLEN;
4530 	(void)iconv(fd, NULL, NULL, &p, &tolen);
4531 	if (p == NULL)
4532 	{
4533 	    iconv_ok = FALSE;
4534 	    iconv_close(fd);
4535 	    fd = (iconv_t)-1;
4536 	}
4537 	else
4538 	    iconv_ok = TRUE;
4539     }
4540 
4541     return (void *)fd;
4542 }
4543 
4544 /*
4545  * Convert the string "str[slen]" with iconv().
4546  * If "unconvlenp" is not NULL handle the string ending in an incomplete
4547  * sequence and set "*unconvlenp" to the length of it.
4548  * Returns the converted string in allocated memory.  NULL for an error.
4549  * If resultlenp is not NULL, sets it to the result length in bytes.
4550  */
4551     static char_u *
4552 iconv_string(
4553     vimconv_T	*vcp,
4554     char_u	*str,
4555     int		slen,
4556     int		*unconvlenp,
4557     int		*resultlenp)
4558 {
4559     const char	*from;
4560     size_t	fromlen;
4561     char	*to;
4562     size_t	tolen;
4563     size_t	len = 0;
4564     size_t	done = 0;
4565     char_u	*result = NULL;
4566     char_u	*p;
4567     int		l;
4568 
4569     from = (char *)str;
4570     fromlen = slen;
4571     for (;;)
4572     {
4573 	if (len == 0 || ICONV_ERRNO == ICONV_E2BIG)
4574 	{
4575 	    /* Allocate enough room for most conversions.  When re-allocating
4576 	     * increase the buffer size. */
4577 	    len = len + fromlen * 2 + 40;
4578 	    p = alloc((unsigned)len);
4579 	    if (p != NULL && done > 0)
4580 		mch_memmove(p, result, done);
4581 	    vim_free(result);
4582 	    result = p;
4583 	    if (result == NULL)	/* out of memory */
4584 		break;
4585 	}
4586 
4587 	to = (char *)result + done;
4588 	tolen = len - done - 2;
4589 	/* Avoid a warning for systems with a wrong iconv() prototype by
4590 	 * casting the second argument to void *. */
4591 	if (iconv(vcp->vc_fd, (void *)&from, &fromlen, &to, &tolen)
4592 								!= (size_t)-1)
4593 	{
4594 	    /* Finished, append a NUL. */
4595 	    *to = NUL;
4596 	    break;
4597 	}
4598 
4599 	/* Check both ICONV_EINVAL and EINVAL, because the dynamically loaded
4600 	 * iconv library may use one of them. */
4601 	if (!vcp->vc_fail && unconvlenp != NULL
4602 		&& (ICONV_ERRNO == ICONV_EINVAL || ICONV_ERRNO == EINVAL))
4603 	{
4604 	    /* Handle an incomplete sequence at the end. */
4605 	    *to = NUL;
4606 	    *unconvlenp = (int)fromlen;
4607 	    break;
4608 	}
4609 
4610 	/* Check both ICONV_EILSEQ and EILSEQ, because the dynamically loaded
4611 	 * iconv library may use one of them. */
4612 	else if (!vcp->vc_fail
4613 		&& (ICONV_ERRNO == ICONV_EILSEQ || ICONV_ERRNO == EILSEQ
4614 		    || ICONV_ERRNO == ICONV_EINVAL || ICONV_ERRNO == EINVAL))
4615 	{
4616 	    /* Can't convert: insert a '?' and skip a character.  This assumes
4617 	     * conversion from 'encoding' to something else.  In other
4618 	     * situations we don't know what to skip anyway. */
4619 	    *to++ = '?';
4620 	    if ((*mb_ptr2cells)((char_u *)from) > 1)
4621 		*to++ = '?';
4622 	    if (enc_utf8)
4623 		l = utfc_ptr2len_len((char_u *)from, (int)fromlen);
4624 	    else
4625 	    {
4626 		l = (*mb_ptr2len)((char_u *)from);
4627 		if (l > (int)fromlen)
4628 		    l = (int)fromlen;
4629 	    }
4630 	    from += l;
4631 	    fromlen -= l;
4632 	}
4633 	else if (ICONV_ERRNO != ICONV_E2BIG)
4634 	{
4635 	    /* conversion failed */
4636 	    VIM_CLEAR(result);
4637 	    break;
4638 	}
4639 	/* Not enough room or skipping illegal sequence. */
4640 	done = to - (char *)result;
4641     }
4642 
4643     if (resultlenp != NULL && result != NULL)
4644 	*resultlenp = (int)(to - (char *)result);
4645     return result;
4646 }
4647 
4648 #  if defined(DYNAMIC_ICONV) || defined(PROTO)
4649 /*
4650  * Dynamically load the "iconv.dll" on Win32.
4651  */
4652 
4653 #   ifndef DYNAMIC_ICONV	    /* must be generating prototypes */
4654 #    define HINSTANCE int
4655 #   endif
4656 static HINSTANCE hIconvDLL = 0;
4657 static HINSTANCE hMsvcrtDLL = 0;
4658 
4659 #   ifndef DYNAMIC_ICONV_DLL
4660 #    define DYNAMIC_ICONV_DLL "iconv.dll"
4661 #    define DYNAMIC_ICONV_DLL_ALT1 "libiconv.dll"
4662 #    define DYNAMIC_ICONV_DLL_ALT2 "libiconv2.dll"
4663 #    define DYNAMIC_ICONV_DLL_ALT3 "libiconv-2.dll"
4664 #   endif
4665 #   ifndef DYNAMIC_MSVCRT_DLL
4666 #    define DYNAMIC_MSVCRT_DLL "msvcrt.dll"
4667 #   endif
4668 
4669 /*
4670  * Try opening the iconv.dll and return TRUE if iconv() can be used.
4671  */
4672     int
4673 iconv_enabled(int verbose)
4674 {
4675     if (hIconvDLL != 0 && hMsvcrtDLL != 0)
4676 	return TRUE;
4677 
4678     /* The iconv DLL file goes under different names, try them all.
4679      * Do the "2" version first, it's newer. */
4680 #ifdef DYNAMIC_ICONV_DLL_ALT2
4681     if (hIconvDLL == 0)
4682 	hIconvDLL = vimLoadLib(DYNAMIC_ICONV_DLL_ALT2);
4683 #endif
4684 #ifdef DYNAMIC_ICONV_DLL_ALT3
4685     if (hIconvDLL == 0)
4686 	hIconvDLL = vimLoadLib(DYNAMIC_ICONV_DLL_ALT3);
4687 #endif
4688     if (hIconvDLL == 0)
4689 	hIconvDLL = vimLoadLib(DYNAMIC_ICONV_DLL);
4690 #ifdef DYNAMIC_ICONV_DLL_ALT1
4691     if (hIconvDLL == 0)
4692 	hIconvDLL = vimLoadLib(DYNAMIC_ICONV_DLL_ALT1);
4693 #endif
4694 
4695     if (hIconvDLL != 0)
4696 	hMsvcrtDLL = vimLoadLib(DYNAMIC_MSVCRT_DLL);
4697     if (hIconvDLL == 0 || hMsvcrtDLL == 0)
4698     {
4699 	/* Only give the message when 'verbose' is set, otherwise it might be
4700 	 * done whenever a conversion is attempted. */
4701 	if (verbose && p_verbose > 0)
4702 	{
4703 	    verbose_enter();
4704 	    semsg(_(e_loadlib),
4705 		    hIconvDLL == 0 ? DYNAMIC_ICONV_DLL : DYNAMIC_MSVCRT_DLL);
4706 	    verbose_leave();
4707 	}
4708 	iconv_end();
4709 	return FALSE;
4710     }
4711 
4712     iconv	= (void *)GetProcAddress(hIconvDLL, "libiconv");
4713     iconv_open	= (void *)GetProcAddress(hIconvDLL, "libiconv_open");
4714     iconv_close	= (void *)GetProcAddress(hIconvDLL, "libiconv_close");
4715     iconvctl	= (void *)GetProcAddress(hIconvDLL, "libiconvctl");
4716     iconv_errno	= get_dll_import_func(hIconvDLL, "_errno");
4717     if (iconv_errno == NULL)
4718 	iconv_errno = (void *)GetProcAddress(hMsvcrtDLL, "_errno");
4719     if (iconv == NULL || iconv_open == NULL || iconv_close == NULL
4720 	    || iconvctl == NULL || iconv_errno == NULL)
4721     {
4722 	iconv_end();
4723 	if (verbose && p_verbose > 0)
4724 	{
4725 	    verbose_enter();
4726 	    semsg(_(e_loadfunc), "for libiconv");
4727 	    verbose_leave();
4728 	}
4729 	return FALSE;
4730     }
4731     return TRUE;
4732 }
4733 
4734     void
4735 iconv_end(void)
4736 {
4737     /* Don't use iconv() when inputting or outputting characters. */
4738     if (input_conv.vc_type == CONV_ICONV)
4739 	convert_setup(&input_conv, NULL, NULL);
4740     if (output_conv.vc_type == CONV_ICONV)
4741 	convert_setup(&output_conv, NULL, NULL);
4742 
4743     if (hIconvDLL != 0)
4744 	FreeLibrary(hIconvDLL);
4745     if (hMsvcrtDLL != 0)
4746 	FreeLibrary(hMsvcrtDLL);
4747     hIconvDLL = 0;
4748     hMsvcrtDLL = 0;
4749 }
4750 #  endif /* DYNAMIC_ICONV */
4751 # endif /* USE_ICONV */
4752 
4753 
4754 #ifdef FEAT_GUI
4755 # define USE_IMACTIVATEFUNC (!gui.in_use && *p_imaf != NUL)
4756 # define USE_IMSTATUSFUNC (!gui.in_use && *p_imsf != NUL)
4757 #else
4758 # define USE_IMACTIVATEFUNC (*p_imaf != NUL)
4759 # define USE_IMSTATUSFUNC (*p_imsf != NUL)
4760 #endif
4761 
4762 #if defined(FEAT_EVAL) && (defined(FEAT_XIM) || defined(IME_WITHOUT_XIM))
4763     static void
4764 call_imactivatefunc(int active)
4765 {
4766     typval_T argv[2];
4767 
4768     argv[0].v_type = VAR_NUMBER;
4769     argv[0].vval.v_number = active ? 1 : 0;
4770     argv[1].v_type = VAR_UNKNOWN;
4771     (void)call_func_retnr(p_imaf, 1, argv);
4772 }
4773 
4774     static int
4775 call_imstatusfunc(void)
4776 {
4777     int is_active;
4778 
4779     /* FIXME: Don't execute user function in unsafe situation. */
4780     if (exiting || is_autocmd_blocked())
4781 	return FALSE;
4782     /* FIXME: :py print 'xxx' is shown duplicate result.
4783      * Use silent to avoid it. */
4784     ++msg_silent;
4785     is_active = call_func_retnr(p_imsf, 0, NULL);
4786     --msg_silent;
4787     return (is_active > 0);
4788 }
4789 #endif
4790 
4791 #if defined(FEAT_XIM) || defined(PROTO)
4792 
4793 # if defined(FEAT_GUI_GTK) || defined(PROTO)
4794 static int xim_has_preediting INIT(= FALSE);  /* IM current status */
4795 
4796 /*
4797  * Set preedit_start_col to the current cursor position.
4798  */
4799     static void
4800 init_preedit_start_col(void)
4801 {
4802     if (State & CMDLINE)
4803 	preedit_start_col = cmdline_getvcol_cursor();
4804     else if (curwin != NULL && curwin->w_buffer != NULL)
4805 	getvcol(curwin, &curwin->w_cursor, &preedit_start_col, NULL, NULL);
4806     /* Prevent that preediting marks the buffer as changed. */
4807     xim_changed_while_preediting = curbuf->b_changed;
4808 }
4809 
4810 static int im_is_active	       = FALSE;	/* IM is enabled for current mode    */
4811 static int preedit_is_active   = FALSE;
4812 static int im_preedit_cursor   = 0;	/* cursor offset in characters       */
4813 static int im_preedit_trailing = 0;	/* number of characters after cursor */
4814 
4815 static unsigned long im_commit_handler_id  = 0;
4816 static unsigned int  im_activatekey_keyval = GDK_VoidSymbol;
4817 static unsigned int  im_activatekey_state  = 0;
4818 
4819 static GtkWidget *preedit_window = NULL;
4820 static GtkWidget *preedit_label = NULL;
4821 
4822 static void im_preedit_window_set_position(void);
4823 
4824     void
4825 im_set_active(int active)
4826 {
4827     int was_active;
4828 
4829     was_active = !!im_get_status();
4830     im_is_active = (active && !p_imdisable);
4831 
4832     if (im_is_active != was_active)
4833 	xim_reset();
4834 }
4835 
4836     void
4837 xim_set_focus(int focus)
4838 {
4839     if (xic != NULL)
4840     {
4841 	if (focus)
4842 	    gtk_im_context_focus_in(xic);
4843 	else
4844 	    gtk_im_context_focus_out(xic);
4845     }
4846 }
4847 
4848     void
4849 im_set_position(int row, int col)
4850 {
4851     if (xic != NULL)
4852     {
4853 	GdkRectangle area;
4854 
4855 	area.x = FILL_X(col);
4856 	area.y = FILL_Y(row);
4857 	area.width  = gui.char_width * (mb_lefthalve(row, col) ? 2 : 1);
4858 	area.height = gui.char_height;
4859 
4860 	gtk_im_context_set_cursor_location(xic, &area);
4861 
4862 	if (p_imst == IM_OVER_THE_SPOT)
4863 	    im_preedit_window_set_position();
4864     }
4865 }
4866 
4867 #  if 0 || defined(PROTO) /* apparently only used in gui_x11.c */
4868     void
4869 xim_set_preedit(void)
4870 {
4871     im_set_position(gui.row, gui.col);
4872 }
4873 #  endif
4874 
4875     static void
4876 im_add_to_input(char_u *str, int len)
4877 {
4878     /* Convert from 'termencoding' (always "utf-8") to 'encoding' */
4879     if (input_conv.vc_type != CONV_NONE)
4880     {
4881 	str = string_convert(&input_conv, str, &len);
4882 	g_return_if_fail(str != NULL);
4883     }
4884 
4885     add_to_input_buf_csi(str, len);
4886 
4887     if (input_conv.vc_type != CONV_NONE)
4888 	vim_free(str);
4889 
4890     if (p_mh) /* blank out the pointer if necessary */
4891 	gui_mch_mousehide(TRUE);
4892 }
4893 
4894      static void
4895 im_preedit_window_set_position(void)
4896 {
4897     int x, y, width, height;
4898     int screen_x, screen_y, screen_width, screen_height;
4899 
4900     if (preedit_window == NULL)
4901 	return;
4902 
4903     gui_gtk_get_screen_geom_of_win(gui.drawarea,
4904 			  &screen_x, &screen_y, &screen_width, &screen_height);
4905     gdk_window_get_origin(gtk_widget_get_window(gui.drawarea), &x, &y);
4906     gtk_window_get_size(GTK_WINDOW(preedit_window), &width, &height);
4907     x = x + FILL_X(gui.col);
4908     y = y + FILL_Y(gui.row);
4909     if (x + width > screen_x + screen_width)
4910 	x = screen_x + screen_width - width;
4911     if (y + height > screen_y + screen_height)
4912 	y = screen_y + screen_height - height;
4913     gtk_window_move(GTK_WINDOW(preedit_window), x, y);
4914 }
4915 
4916     static void
4917 im_preedit_window_open()
4918 {
4919     char *preedit_string;
4920 #if !GTK_CHECK_VERSION(3,16,0)
4921     char buf[8];
4922 #endif
4923     PangoAttrList *attr_list;
4924     PangoLayout *layout;
4925 #if GTK_CHECK_VERSION(3,0,0)
4926 # if !GTK_CHECK_VERSION(3,16,0)
4927     GdkRGBA color;
4928 # endif
4929 #else
4930     GdkColor color;
4931 #endif
4932     gint w, h;
4933 
4934     if (preedit_window == NULL)
4935     {
4936 	preedit_window = gtk_window_new(GTK_WINDOW_POPUP);
4937 	gtk_window_set_transient_for(GTK_WINDOW(preedit_window),
4938 						     GTK_WINDOW(gui.mainwin));
4939 	preedit_label = gtk_label_new("");
4940 	gtk_widget_set_name(preedit_label, "vim-gui-preedit-area");
4941 	gtk_container_add(GTK_CONTAINER(preedit_window), preedit_label);
4942     }
4943 
4944 #if GTK_CHECK_VERSION(3,16,0)
4945     {
4946 	GtkStyleContext * const context
4947 				  = gtk_widget_get_style_context(gui.drawarea);
4948 	GtkCssProvider * const provider = gtk_css_provider_new();
4949 	gchar		   *css = NULL;
4950 	const char * const fontname
4951 			   = pango_font_description_get_family(gui.norm_font);
4952 	gint	fontsize
4953 		= pango_font_description_get_size(gui.norm_font) / PANGO_SCALE;
4954 	gchar	*fontsize_propval = NULL;
4955 
4956 	if (!pango_font_description_get_size_is_absolute(gui.norm_font))
4957 	{
4958 	    /* fontsize was given in points.  Convert it into that in pixels
4959 	     * to use with CSS. */
4960 	    GdkScreen * const screen
4961 		  = gdk_window_get_screen(gtk_widget_get_window(gui.mainwin));
4962 	    const gdouble dpi = gdk_screen_get_resolution(screen);
4963 	    fontsize = dpi * fontsize / 72;
4964 	}
4965 	if (fontsize > 0)
4966 	    fontsize_propval = g_strdup_printf("%dpx", fontsize);
4967 	else
4968 	    fontsize_propval = g_strdup_printf("inherit");
4969 
4970 	css = g_strdup_printf(
4971 		"widget#vim-gui-preedit-area {\n"
4972 		"  font-family: %s,monospace;\n"
4973 		"  font-size: %s;\n"
4974 		"  color: #%.2lx%.2lx%.2lx;\n"
4975 		"  background-color: #%.2lx%.2lx%.2lx;\n"
4976 		"}\n",
4977 		fontname != NULL ? fontname : "inherit",
4978 		fontsize_propval,
4979 		(gui.norm_pixel >> 16) & 0xff,
4980 		(gui.norm_pixel >> 8) & 0xff,
4981 		gui.norm_pixel & 0xff,
4982 		(gui.back_pixel >> 16) & 0xff,
4983 		(gui.back_pixel >> 8) & 0xff,
4984 		gui.back_pixel & 0xff);
4985 
4986 	gtk_css_provider_load_from_data(provider, css, -1, NULL);
4987 	gtk_style_context_add_provider(context,
4988 				     GTK_STYLE_PROVIDER(provider), G_MAXUINT);
4989 
4990 	g_free(css);
4991 	g_free(fontsize_propval);
4992 	g_object_unref(provider);
4993     }
4994 #elif GTK_CHECK_VERSION(3,0,0)
4995     gtk_widget_override_font(preedit_label, gui.norm_font);
4996 
4997     vim_snprintf(buf, sizeof(buf), "#%06X", gui.norm_pixel);
4998     gdk_rgba_parse(&color, buf);
4999     gtk_widget_override_color(preedit_label, GTK_STATE_FLAG_NORMAL, &color);
5000 
5001     vim_snprintf(buf, sizeof(buf), "#%06X", gui.back_pixel);
5002     gdk_rgba_parse(&color, buf);
5003     gtk_widget_override_background_color(preedit_label, GTK_STATE_FLAG_NORMAL,
5004 								      &color);
5005 #else
5006     gtk_widget_modify_font(preedit_label, gui.norm_font);
5007 
5008     vim_snprintf(buf, sizeof(buf), "#%06X", (unsigned)gui.norm_pixel);
5009     gdk_color_parse(buf, &color);
5010     gtk_widget_modify_fg(preedit_label, GTK_STATE_NORMAL, &color);
5011 
5012     vim_snprintf(buf, sizeof(buf), "#%06X", (unsigned)gui.back_pixel);
5013     gdk_color_parse(buf, &color);
5014     gtk_widget_modify_bg(preedit_window, GTK_STATE_NORMAL, &color);
5015 #endif
5016 
5017     gtk_im_context_get_preedit_string(xic, &preedit_string, &attr_list, NULL);
5018 
5019     if (preedit_string[0] != NUL)
5020     {
5021 	gtk_label_set_text(GTK_LABEL(preedit_label), preedit_string);
5022 	gtk_label_set_attributes(GTK_LABEL(preedit_label), attr_list);
5023 
5024 	layout = gtk_label_get_layout(GTK_LABEL(preedit_label));
5025 	pango_layout_get_pixel_size(layout, &w, &h);
5026 	h = MAX(h, gui.char_height);
5027 	gtk_window_resize(GTK_WINDOW(preedit_window), w, h);
5028 
5029 	gtk_widget_show_all(preedit_window);
5030 
5031 	im_preedit_window_set_position();
5032     }
5033 
5034     g_free(preedit_string);
5035     pango_attr_list_unref(attr_list);
5036 }
5037 
5038     static void
5039 im_preedit_window_close()
5040 {
5041     if (preedit_window != NULL)
5042 	gtk_widget_hide(preedit_window);
5043 }
5044 
5045     static void
5046 im_show_preedit()
5047 {
5048     im_preedit_window_open();
5049 
5050     if (p_mh) /* blank out the pointer if necessary */
5051 	gui_mch_mousehide(TRUE);
5052 }
5053 
5054     static void
5055 im_delete_preedit(void)
5056 {
5057     char_u bskey[]  = {CSI, 'k', 'b'};
5058     char_u delkey[] = {CSI, 'k', 'D'};
5059 
5060     if (p_imst == IM_OVER_THE_SPOT)
5061     {
5062 	im_preedit_window_close();
5063 	return;
5064     }
5065 
5066     if (State & NORMAL
5067 #ifdef FEAT_TERMINAL
5068 	    && !term_use_loop()
5069 #endif
5070        )
5071     {
5072 	im_preedit_cursor = 0;
5073 	return;
5074     }
5075     for (; im_preedit_cursor > 0; --im_preedit_cursor)
5076 	add_to_input_buf(bskey, (int)sizeof(bskey));
5077 
5078     for (; im_preedit_trailing > 0; --im_preedit_trailing)
5079 	add_to_input_buf(delkey, (int)sizeof(delkey));
5080 }
5081 
5082 /*
5083  * Move the cursor left by "num_move_back" characters.
5084  * Note that ins_left() checks im_is_preediting() to avoid breaking undo for
5085  * these K_LEFT keys.
5086  */
5087     static void
5088 im_correct_cursor(int num_move_back)
5089 {
5090     char_u backkey[] = {CSI, 'k', 'l'};
5091 
5092     if (State & NORMAL)
5093 	return;
5094 #  ifdef FEAT_RIGHTLEFT
5095     if ((State & CMDLINE) == 0 && curwin != NULL && curwin->w_p_rl)
5096 	backkey[2] = 'r';
5097 #  endif
5098     for (; num_move_back > 0; --num_move_back)
5099 	add_to_input_buf(backkey, (int)sizeof(backkey));
5100 }
5101 
5102 static int xim_expected_char = NUL;
5103 static int xim_ignored_char = FALSE;
5104 
5105 /*
5106  * Update the mode and cursor while in an IM callback.
5107  */
5108     static void
5109 im_show_info(void)
5110 {
5111     int	    old_vgetc_busy;
5112 
5113     old_vgetc_busy = vgetc_busy;
5114     vgetc_busy = TRUE;
5115     showmode();
5116     vgetc_busy = old_vgetc_busy;
5117     if ((State & NORMAL) || (State & INSERT))
5118 	setcursor();
5119     out_flush();
5120 }
5121 
5122 /*
5123  * Callback invoked when the user finished preediting.
5124  * Put the final string into the input buffer.
5125  */
5126     static void
5127 im_commit_cb(GtkIMContext *context UNUSED,
5128 	     const gchar *str,
5129 	     gpointer data UNUSED)
5130 {
5131     int		slen = (int)STRLEN(str);
5132     int		add_to_input = TRUE;
5133     int		clen;
5134     int		len = slen;
5135     int		commit_with_preedit = TRUE;
5136     char_u	*im_str;
5137 
5138 #ifdef XIM_DEBUG
5139     xim_log("im_commit_cb(): %s\n", str);
5140 #endif
5141 
5142     if (p_imst == IM_ON_THE_SPOT)
5143     {
5144 	/* The imhangul module doesn't reset the preedit string before
5145 	 * committing.  Call im_delete_preedit() to work around that. */
5146 	im_delete_preedit();
5147 
5148 	/* Indicate that preediting has finished. */
5149 	if (preedit_start_col == MAXCOL)
5150 	{
5151 	    init_preedit_start_col();
5152 	    commit_with_preedit = FALSE;
5153 	}
5154 
5155 	/* The thing which setting "preedit_start_col" to MAXCOL means that
5156 	 * "preedit_start_col" will be set forcedly when calling
5157 	 * preedit_changed_cb() next time.
5158 	 * "preedit_start_col" should not reset with MAXCOL on this part. Vim
5159 	 * is simulating the preediting by using add_to_input_str(). when
5160 	 * preedit begin immediately before committed, the typebuf is not
5161 	 * flushed to screen, then it can't get correct "preedit_start_col".
5162 	 * Thus, it should calculate the cells by adding cells of the committed
5163 	 * string. */
5164 	if (input_conv.vc_type != CONV_NONE)
5165 	{
5166 	    im_str = string_convert(&input_conv, (char_u *)str, &len);
5167 	    g_return_if_fail(im_str != NULL);
5168 	}
5169 	else
5170 	    im_str = (char_u *)str;
5171 
5172 	clen = mb_string2cells(im_str, len);
5173 
5174 	if (input_conv.vc_type != CONV_NONE)
5175 	    vim_free(im_str);
5176 	preedit_start_col += clen;
5177     }
5178 
5179     /* Is this a single character that matches a keypad key that's just
5180      * been pressed?  If so, we don't want it to be entered as such - let
5181      * us carry on processing the raw keycode so that it may be used in
5182      * mappings as <kSomething>. */
5183     if (xim_expected_char != NUL)
5184     {
5185 	/* We're currently processing a keypad or other special key */
5186 	if (slen == 1 && str[0] == xim_expected_char)
5187 	{
5188 	    /* It's a match - don't do it here */
5189 	    xim_ignored_char = TRUE;
5190 	    add_to_input = FALSE;
5191 	}
5192 	else
5193 	{
5194 	    /* Not a match */
5195 	    xim_ignored_char = FALSE;
5196 	}
5197     }
5198 
5199     if (add_to_input)
5200 	im_add_to_input((char_u *)str, slen);
5201 
5202     if (p_imst == IM_ON_THE_SPOT)
5203     {
5204 	/* Inserting chars while "im_is_active" is set does not cause a
5205 	 * change of buffer.  When the chars are committed the buffer must be
5206 	 * marked as changed. */
5207 	if (!commit_with_preedit)
5208 	    preedit_start_col = MAXCOL;
5209 
5210 	/* This flag is used in changed() at next call. */
5211 	xim_changed_while_preediting = TRUE;
5212     }
5213 
5214     if (gtk_main_level() > 0)
5215 	gtk_main_quit();
5216 }
5217 
5218 /*
5219  * Callback invoked after start to the preedit.
5220  */
5221     static void
5222 im_preedit_start_cb(GtkIMContext *context UNUSED, gpointer data UNUSED)
5223 {
5224 #ifdef XIM_DEBUG
5225     xim_log("im_preedit_start_cb()\n");
5226 #endif
5227 
5228     im_is_active = TRUE;
5229     preedit_is_active = TRUE;
5230     gui_update_cursor(TRUE, FALSE);
5231     im_show_info();
5232 }
5233 
5234 /*
5235  * Callback invoked after end to the preedit.
5236  */
5237     static void
5238 im_preedit_end_cb(GtkIMContext *context UNUSED, gpointer data UNUSED)
5239 {
5240 #ifdef XIM_DEBUG
5241     xim_log("im_preedit_end_cb()\n");
5242 #endif
5243     im_delete_preedit();
5244 
5245     /* Indicate that preediting has finished */
5246     if (p_imst == IM_ON_THE_SPOT)
5247 	preedit_start_col = MAXCOL;
5248     xim_has_preediting = FALSE;
5249 
5250 #if 0
5251     /* Removal of this line suggested by Takuhiro Nishioka.  Fixes that IM was
5252      * switched off unintentionally.  We now use preedit_is_active (added by
5253      * SungHyun Nam). */
5254     im_is_active = FALSE;
5255 #endif
5256     preedit_is_active = FALSE;
5257     gui_update_cursor(TRUE, FALSE);
5258     im_show_info();
5259 }
5260 
5261 /*
5262  * Callback invoked after changes to the preedit string.  If the preedit
5263  * string was empty before, remember the preedit start column so we know
5264  * where to apply feedback attributes.  Delete the previous preedit string
5265  * if there was one, save the new preedit cursor offset, and put the new
5266  * string into the input buffer.
5267  *
5268  * TODO: The pragmatic "put into input buffer" approach used here has
5269  *       several fundamental problems:
5270  *
5271  * - The characters in the preedit string are subject to remapping.
5272  *   That's broken, only the finally committed string should be remapped.
5273  *
5274  * - There is a race condition involved:  The retrieved value for the
5275  *   current cursor position will be wrong if any unprocessed characters
5276  *   are still queued in the input buffer.
5277  *
5278  * - Due to the lack of synchronization between the file buffer in memory
5279  *   and any typed characters, it's practically impossible to implement the
5280  *   "retrieve_surrounding" and "delete_surrounding" signals reliably.  IM
5281  *   modules for languages such as Thai are likely to rely on this feature
5282  *   for proper operation.
5283  *
5284  * Conclusions:  I think support for preediting needs to be moved to the
5285  * core parts of Vim.  Ideally, until it has been committed, the preediting
5286  * string should only be displayed and not affect the buffer content at all.
5287  * The question how to deal with the synchronization issue still remains.
5288  * Circumventing the input buffer is probably not desirable.  Anyway, I think
5289  * implementing "retrieve_surrounding" is the only hard problem.
5290  *
5291  * One way to solve all of this in a clean manner would be to queue all key
5292  * press/release events "as is" in the input buffer, and apply the IM filtering
5293  * at the receiving end of the queue.  This, however, would have a rather large
5294  * impact on the code base.  If there is an easy way to force processing of all
5295  * remaining input from within the "retrieve_surrounding" signal handler, this
5296  * might not be necessary.  Gotta ask on vim-dev for opinions.
5297  */
5298     static void
5299 im_preedit_changed_cb(GtkIMContext *context, gpointer data UNUSED)
5300 {
5301     char    *preedit_string = NULL;
5302     int	    cursor_index    = 0;
5303     int	    num_move_back   = 0;
5304     char_u  *str;
5305     char_u  *p;
5306     int	    i;
5307 
5308     if (p_imst == IM_ON_THE_SPOT)
5309 	gtk_im_context_get_preedit_string(context,
5310 					  &preedit_string, NULL,
5311 					  &cursor_index);
5312     else
5313 	gtk_im_context_get_preedit_string(context,
5314 					  &preedit_string, NULL,
5315 					  NULL);
5316 
5317 #ifdef XIM_DEBUG
5318     xim_log("im_preedit_changed_cb(): %s\n", preedit_string);
5319 #endif
5320 
5321     g_return_if_fail(preedit_string != NULL); /* just in case */
5322 
5323     if (p_imst == IM_OVER_THE_SPOT)
5324     {
5325 	if (preedit_string[0] == NUL)
5326 	{
5327 	    xim_has_preediting = FALSE;
5328 	    im_delete_preedit();
5329 	}
5330 	else
5331 	{
5332 	    xim_has_preediting = TRUE;
5333 	    im_show_preedit();
5334 	}
5335     }
5336     else
5337     {
5338 	/* If preedit_start_col is MAXCOL set it to the current cursor position. */
5339 	if (preedit_start_col == MAXCOL && preedit_string[0] != '\0')
5340 	{
5341 	    xim_has_preediting = TRUE;
5342 
5343 	    /* Urgh, this breaks if the input buffer isn't empty now */
5344 	    init_preedit_start_col();
5345 	}
5346 	else if (cursor_index == 0 && preedit_string[0] == '\0')
5347 	{
5348 	    xim_has_preediting = FALSE;
5349 
5350 	    /* If at the start position (after typing backspace)
5351 	     * preedit_start_col must be reset. */
5352 	    preedit_start_col = MAXCOL;
5353 	}
5354 
5355 	im_delete_preedit();
5356 
5357 	/*
5358 	 * Compute the end of the preediting area: "preedit_end_col".
5359 	 * According to the documentation of gtk_im_context_get_preedit_string(),
5360 	 * the cursor_pos output argument returns the offset in bytes.  This is
5361 	 * unfortunately not true -- real life shows the offset is in characters,
5362 	 * and the GTK+ source code agrees with me.  Will file a bug later.
5363 	 */
5364 	if (preedit_start_col != MAXCOL)
5365 	    preedit_end_col = preedit_start_col;
5366 	str = (char_u *)preedit_string;
5367 	for (p = str, i = 0; *p != NUL; p += utf_byte2len(*p), ++i)
5368 	{
5369 	    int is_composing;
5370 
5371 	    is_composing = ((*p & 0x80) != 0 && utf_iscomposing(utf_ptr2char(p)));
5372 	    /*
5373 	     * These offsets are used as counters when generating <BS> and <Del>
5374 	     * to delete the preedit string.  So don't count composing characters
5375 	     * unless 'delcombine' is enabled.
5376 	     */
5377 	    if (!is_composing || p_deco)
5378 	    {
5379 		if (i < cursor_index)
5380 		    ++im_preedit_cursor;
5381 		else
5382 		    ++im_preedit_trailing;
5383 	    }
5384 	    if (!is_composing && i >= cursor_index)
5385 	    {
5386 		/* This is essentially the same as im_preedit_trailing, except
5387 		 * composing characters are not counted even if p_deco is set. */
5388 		++num_move_back;
5389 	    }
5390 	    if (preedit_start_col != MAXCOL)
5391 		preedit_end_col += utf_ptr2cells(p);
5392 	}
5393 
5394 	if (p > str)
5395 	{
5396 	    im_add_to_input(str, (int)(p - str));
5397 	    im_correct_cursor(num_move_back);
5398 	}
5399     }
5400 
5401     g_free(preedit_string);
5402 
5403     if (gtk_main_level() > 0)
5404 	gtk_main_quit();
5405 }
5406 
5407 /*
5408  * Translate the Pango attributes at iter to Vim highlighting attributes.
5409  * Ignore attributes not supported by Vim highlighting.  This shouldn't have
5410  * too much impact -- right now we handle even more attributes than necessary
5411  * for the IM modules I tested with.
5412  */
5413     static int
5414 translate_pango_attributes(PangoAttrIterator *iter)
5415 {
5416     PangoAttribute  *attr;
5417     int		    char_attr = HL_NORMAL;
5418 
5419     attr = pango_attr_iterator_get(iter, PANGO_ATTR_UNDERLINE);
5420     if (attr != NULL && ((PangoAttrInt *)attr)->value
5421 						 != (int)PANGO_UNDERLINE_NONE)
5422 	char_attr |= HL_UNDERLINE;
5423 
5424     attr = pango_attr_iterator_get(iter, PANGO_ATTR_WEIGHT);
5425     if (attr != NULL && ((PangoAttrInt *)attr)->value >= (int)PANGO_WEIGHT_BOLD)
5426 	char_attr |= HL_BOLD;
5427 
5428     attr = pango_attr_iterator_get(iter, PANGO_ATTR_STYLE);
5429     if (attr != NULL && ((PangoAttrInt *)attr)->value
5430 						   != (int)PANGO_STYLE_NORMAL)
5431 	char_attr |= HL_ITALIC;
5432 
5433     attr = pango_attr_iterator_get(iter, PANGO_ATTR_BACKGROUND);
5434     if (attr != NULL)
5435     {
5436 	const PangoColor *color = &((PangoAttrColor *)attr)->color;
5437 
5438 	/* Assume inverse if black background is requested */
5439 	if ((color->red | color->green | color->blue) == 0)
5440 	    char_attr |= HL_INVERSE;
5441     }
5442 
5443     return char_attr;
5444 }
5445 
5446 /*
5447  * Retrieve the highlighting attributes at column col in the preedit string.
5448  * Return -1 if not in preediting mode or if col is out of range.
5449  */
5450     int
5451 im_get_feedback_attr(int col)
5452 {
5453     char	    *preedit_string = NULL;
5454     PangoAttrList   *attr_list	    = NULL;
5455     int		    char_attr	    = -1;
5456 
5457     if (xic == NULL)
5458 	return char_attr;
5459 
5460     gtk_im_context_get_preedit_string(xic, &preedit_string, &attr_list, NULL);
5461 
5462     if (preedit_string != NULL && attr_list != NULL)
5463     {
5464 	int idx;
5465 
5466 	/* Get the byte index as used by PangoAttrIterator */
5467 	for (idx = 0; col > 0 && preedit_string[idx] != '\0'; --col)
5468 	    idx += utfc_ptr2len((char_u *)preedit_string + idx);
5469 
5470 	if (preedit_string[idx] != '\0')
5471 	{
5472 	    PangoAttrIterator	*iter;
5473 	    int			start, end;
5474 
5475 	    char_attr = HL_NORMAL;
5476 	    iter = pango_attr_list_get_iterator(attr_list);
5477 
5478 	    /* Extract all relevant attributes from the list. */
5479 	    do
5480 	    {
5481 		pango_attr_iterator_range(iter, &start, &end);
5482 
5483 		if (idx >= start && idx < end)
5484 		    char_attr |= translate_pango_attributes(iter);
5485 	    }
5486 	    while (pango_attr_iterator_next(iter));
5487 
5488 	    pango_attr_iterator_destroy(iter);
5489 	}
5490     }
5491 
5492     if (attr_list != NULL)
5493 	pango_attr_list_unref(attr_list);
5494     g_free(preedit_string);
5495 
5496     return char_attr;
5497 }
5498 
5499     void
5500 xim_init(void)
5501 {
5502 #ifdef XIM_DEBUG
5503     xim_log("xim_init()\n");
5504 #endif
5505 
5506     g_return_if_fail(gui.drawarea != NULL);
5507     g_return_if_fail(gtk_widget_get_window(gui.drawarea) != NULL);
5508 
5509     xic = gtk_im_multicontext_new();
5510     g_object_ref(xic);
5511 
5512     im_commit_handler_id = g_signal_connect(G_OBJECT(xic), "commit",
5513 					    G_CALLBACK(&im_commit_cb), NULL);
5514     g_signal_connect(G_OBJECT(xic), "preedit_changed",
5515 		     G_CALLBACK(&im_preedit_changed_cb), NULL);
5516     g_signal_connect(G_OBJECT(xic), "preedit_start",
5517 		     G_CALLBACK(&im_preedit_start_cb), NULL);
5518     g_signal_connect(G_OBJECT(xic), "preedit_end",
5519 		     G_CALLBACK(&im_preedit_end_cb), NULL);
5520 
5521     gtk_im_context_set_client_window(xic, gtk_widget_get_window(gui.drawarea));
5522 }
5523 
5524     void
5525 im_shutdown(void)
5526 {
5527 #ifdef XIM_DEBUG
5528     xim_log("im_shutdown()\n");
5529 #endif
5530 
5531     if (xic != NULL)
5532     {
5533 	gtk_im_context_focus_out(xic);
5534 	g_object_unref(xic);
5535 	xic = NULL;
5536     }
5537     im_is_active = FALSE;
5538     im_commit_handler_id = 0;
5539     if (p_imst == IM_ON_THE_SPOT)
5540 	preedit_start_col = MAXCOL;
5541     xim_has_preediting = FALSE;
5542 }
5543 
5544 /*
5545  * Convert the string argument to keyval and state for GdkEventKey.
5546  * If str is valid return TRUE, otherwise FALSE.
5547  *
5548  * See 'imactivatekey' for documentation of the format.
5549  */
5550     static int
5551 im_string_to_keyval(const char *str, unsigned int *keyval, unsigned int *state)
5552 {
5553     const char	    *mods_end;
5554     unsigned	    tmp_keyval;
5555     unsigned	    tmp_state = 0;
5556 
5557     mods_end = strrchr(str, '-');
5558     mods_end = (mods_end != NULL) ? mods_end + 1 : str;
5559 
5560     /* Parse modifier keys */
5561     while (str < mods_end)
5562 	switch (*str++)
5563 	{
5564 	    case '-':							break;
5565 	    case 'S': case 's': tmp_state |= (unsigned)GDK_SHIFT_MASK;	break;
5566 	    case 'L': case 'l': tmp_state |= (unsigned)GDK_LOCK_MASK;	break;
5567 	    case 'C': case 'c': tmp_state |= (unsigned)GDK_CONTROL_MASK;break;
5568 	    case '1':		tmp_state |= (unsigned)GDK_MOD1_MASK;	break;
5569 	    case '2':		tmp_state |= (unsigned)GDK_MOD2_MASK;	break;
5570 	    case '3':		tmp_state |= (unsigned)GDK_MOD3_MASK;	break;
5571 	    case '4':		tmp_state |= (unsigned)GDK_MOD4_MASK;	break;
5572 	    case '5':		tmp_state |= (unsigned)GDK_MOD5_MASK;	break;
5573 	    default:
5574 		return FALSE;
5575 	}
5576 
5577     tmp_keyval = gdk_keyval_from_name(str);
5578 
5579     if (tmp_keyval == 0 || tmp_keyval == GDK_VoidSymbol)
5580 	return FALSE;
5581 
5582     if (keyval != NULL)
5583 	*keyval = tmp_keyval;
5584     if (state != NULL)
5585 	*state = tmp_state;
5586 
5587     return TRUE;
5588 }
5589 
5590 /*
5591  * Return TRUE if p_imak is valid, otherwise FALSE.  As a special case, an
5592  * empty string is also regarded as valid.
5593  *
5594  * Note: The numerical key value of p_imak is cached if it was valid; thus
5595  * boldly assuming im_xim_isvalid_imactivate() will always be called whenever
5596  * 'imak' changes.  This is currently the case but not obvious -- should
5597  * probably rename the function for clarity.
5598  */
5599     int
5600 im_xim_isvalid_imactivate(void)
5601 {
5602     if (p_imak[0] == NUL)
5603     {
5604 	im_activatekey_keyval = GDK_VoidSymbol;
5605 	im_activatekey_state  = 0;
5606 	return TRUE;
5607     }
5608 
5609     return im_string_to_keyval((const char *)p_imak,
5610 			       &im_activatekey_keyval,
5611 			       &im_activatekey_state);
5612 }
5613 
5614     static void
5615 im_synthesize_keypress(unsigned int keyval, unsigned int state)
5616 {
5617     GdkEventKey *event;
5618 
5619     event = (GdkEventKey *)gdk_event_new(GDK_KEY_PRESS);
5620     g_object_ref(gtk_widget_get_window(gui.drawarea));
5621 					/* unreffed by gdk_event_free() */
5622     event->window = gtk_widget_get_window(gui.drawarea);
5623     event->send_event = TRUE;
5624     event->time = GDK_CURRENT_TIME;
5625     event->state  = state;
5626     event->keyval = keyval;
5627     event->hardware_keycode = /* needed for XIM */
5628 	XKeysymToKeycode(GDK_WINDOW_XDISPLAY(event->window), (KeySym)keyval);
5629     event->length = 0;
5630     event->string = NULL;
5631 
5632     gtk_im_context_filter_keypress(xic, event);
5633 
5634     /* For consistency, also send the corresponding release event. */
5635     event->type = GDK_KEY_RELEASE;
5636     event->send_event = FALSE;
5637     gtk_im_context_filter_keypress(xic, event);
5638 
5639     gdk_event_free((GdkEvent *)event);
5640 }
5641 
5642     void
5643 xim_reset(void)
5644 {
5645 # ifdef FEAT_EVAL
5646     if (USE_IMACTIVATEFUNC)
5647 	call_imactivatefunc(im_is_active);
5648     else
5649 # endif
5650     if (xic != NULL)
5651     {
5652 	gtk_im_context_reset(xic);
5653 
5654 	if (p_imdisable)
5655 	    im_shutdown();
5656 	else
5657 	{
5658 	    xim_set_focus(gui.in_focus);
5659 
5660 	    if (im_activatekey_keyval != GDK_VoidSymbol)
5661 	    {
5662 		if (im_is_active)
5663 		{
5664 		    g_signal_handler_block(xic, im_commit_handler_id);
5665 		    im_synthesize_keypress(im_activatekey_keyval,
5666 						    im_activatekey_state);
5667 		    g_signal_handler_unblock(xic, im_commit_handler_id);
5668 		}
5669 	    }
5670 	    else
5671 	    {
5672 		im_shutdown();
5673 		xim_init();
5674 		xim_set_focus(gui.in_focus);
5675 	    }
5676 	}
5677     }
5678 
5679     if (p_imst == IM_ON_THE_SPOT)
5680 	preedit_start_col = MAXCOL;
5681     xim_has_preediting = FALSE;
5682 }
5683 
5684     int
5685 xim_queue_key_press_event(GdkEventKey *event, int down)
5686 {
5687     if (down)
5688     {
5689 	/*
5690 	 * Workaround GTK2 XIM 'feature' that always converts keypad keys to
5691 	 * chars., even when not part of an IM sequence (ref. feature of
5692 	 * gdk/gdkkeyuni.c).
5693 	 * Flag any keypad keys that might represent a single char.
5694 	 * If this (on its own - i.e., not part of an IM sequence) is
5695 	 * committed while we're processing one of these keys, we can ignore
5696 	 * that commit and go ahead & process it ourselves.  That way we can
5697 	 * still distinguish keypad keys for use in mappings.
5698 	 * Also add GDK_space to make <S-Space> work.
5699 	 */
5700 	switch (event->keyval)
5701 	{
5702 	    case GDK_KP_Add:      xim_expected_char = '+';  break;
5703 	    case GDK_KP_Subtract: xim_expected_char = '-';  break;
5704 	    case GDK_KP_Divide:   xim_expected_char = '/';  break;
5705 	    case GDK_KP_Multiply: xim_expected_char = '*';  break;
5706 	    case GDK_KP_Decimal:  xim_expected_char = '.';  break;
5707 	    case GDK_KP_Equal:    xim_expected_char = '=';  break;
5708 	    case GDK_KP_0:	  xim_expected_char = '0';  break;
5709 	    case GDK_KP_1:	  xim_expected_char = '1';  break;
5710 	    case GDK_KP_2:	  xim_expected_char = '2';  break;
5711 	    case GDK_KP_3:	  xim_expected_char = '3';  break;
5712 	    case GDK_KP_4:	  xim_expected_char = '4';  break;
5713 	    case GDK_KP_5:	  xim_expected_char = '5';  break;
5714 	    case GDK_KP_6:	  xim_expected_char = '6';  break;
5715 	    case GDK_KP_7:	  xim_expected_char = '7';  break;
5716 	    case GDK_KP_8:	  xim_expected_char = '8';  break;
5717 	    case GDK_KP_9:	  xim_expected_char = '9';  break;
5718 	    case GDK_space:	  xim_expected_char = ' ';  break;
5719 	    default:		  xim_expected_char = NUL;
5720 	}
5721 	xim_ignored_char = FALSE;
5722     }
5723 
5724     /*
5725      * When typing fFtT, XIM may be activated. Thus it must pass
5726      * gtk_im_context_filter_keypress() in Normal mode.
5727      * And while doing :sh too.
5728      */
5729     if (xic != NULL && !p_imdisable
5730 		    && (State & (INSERT | CMDLINE | NORMAL | EXTERNCMD)) != 0)
5731     {
5732 	/*
5733 	 * Filter 'imactivatekey' and map it to CTRL-^.  This way, Vim is
5734 	 * always aware of the current status of IM, and can even emulate
5735 	 * the activation key for modules that don't support one.
5736 	 */
5737 	if (event->keyval == im_activatekey_keyval
5738 	     && (event->state & im_activatekey_state) == im_activatekey_state)
5739 	{
5740 	    unsigned int state_mask;
5741 
5742 	    /* Require the state of the 3 most used modifiers to match exactly.
5743 	     * Otherwise e.g. <S-C-space> would be unusable for other purposes
5744 	     * if the IM activate key is <S-space>. */
5745 	    state_mask  = im_activatekey_state;
5746 	    state_mask |= ((int)GDK_SHIFT_MASK | (int)GDK_CONTROL_MASK
5747 							| (int)GDK_MOD1_MASK);
5748 
5749 	    if ((event->state & state_mask) != im_activatekey_state)
5750 		return FALSE;
5751 
5752 	    /* Don't send it a second time on GDK_KEY_RELEASE. */
5753 	    if (event->type != GDK_KEY_PRESS)
5754 		return TRUE;
5755 
5756 	    if (map_to_exists_mode((char_u *)"", LANGMAP, FALSE))
5757 	    {
5758 		im_set_active(FALSE);
5759 
5760 		/* ":lmap" mappings exists, toggle use of mappings. */
5761 		State ^= LANGMAP;
5762 		if (State & LANGMAP)
5763 		{
5764 		    curbuf->b_p_iminsert = B_IMODE_NONE;
5765 		    State &= ~LANGMAP;
5766 		}
5767 		else
5768 		{
5769 		    curbuf->b_p_iminsert = B_IMODE_LMAP;
5770 		    State |= LANGMAP;
5771 		}
5772 		return TRUE;
5773 	    }
5774 
5775 	    return gtk_im_context_filter_keypress(xic, event);
5776 	}
5777 
5778 	/* Don't filter events through the IM context if IM isn't active
5779 	 * right now.  Unlike with GTK+ 1.2 we cannot rely on the IM module
5780 	 * not doing anything before the activation key was sent. */
5781 	if (im_activatekey_keyval == GDK_VoidSymbol || im_is_active)
5782 	{
5783 	    int imresult = gtk_im_context_filter_keypress(xic, event);
5784 
5785 	    if (p_imst == IM_ON_THE_SPOT)
5786 	    {
5787 		/* Some XIM send following sequence:
5788 		 * 1. preedited string.
5789 		 * 2. committed string.
5790 		 * 3. line changed key.
5791 		 * 4. preedited string.
5792 		 * 5. remove preedited string.
5793 		 * if 3, Vim can't move back the above line for 5.
5794 		 * thus, this part should not parse the key. */
5795 		if (!imresult && preedit_start_col != MAXCOL
5796 					    && event->keyval == GDK_Return)
5797 		{
5798 		    im_synthesize_keypress(GDK_Return, 0U);
5799 		    return FALSE;
5800 		}
5801 	    }
5802 
5803 	    /* If XIM tried to commit a keypad key as a single char.,
5804 	     * ignore it so we can use the keypad key 'raw', for mappings. */
5805 	    if (xim_expected_char != NUL && xim_ignored_char)
5806 		/* We had a keypad key, and XIM tried to thieve it */
5807 		return FALSE;
5808 
5809 	    /* This is supposed to fix a problem with iBus, that space
5810 	     * characters don't work in input mode. */
5811 	    xim_expected_char = NUL;
5812 
5813 	    /* Normal processing */
5814 	    return imresult;
5815 	}
5816     }
5817 
5818     return FALSE;
5819 }
5820 
5821     int
5822 im_get_status(void)
5823 {
5824 #  ifdef FEAT_EVAL
5825     if (USE_IMSTATUSFUNC)
5826 	return call_imstatusfunc();
5827 #  endif
5828     return im_is_active;
5829 }
5830 
5831     int
5832 preedit_get_status(void)
5833 {
5834     return preedit_is_active;
5835 }
5836 
5837     int
5838 im_is_preediting(void)
5839 {
5840     return xim_has_preediting;
5841 }
5842 
5843 # else /* !FEAT_GUI_GTK */
5844 
5845 static int	xim_is_active = FALSE;  /* XIM should be active in the current
5846 					   mode */
5847 static int	xim_has_focus = FALSE;	/* XIM is really being used for Vim */
5848 #  ifdef FEAT_GUI_X11
5849 static XIMStyle	input_style;
5850 static int	status_area_enabled = TRUE;
5851 #  endif
5852 
5853 /*
5854  * Switch using XIM on/off.  This is used by the code that changes "State".
5855  * When 'imactivatefunc' is defined use that function instead.
5856  */
5857     void
5858 im_set_active(int active_arg)
5859 {
5860     int active = active_arg;
5861 
5862     /* If 'imdisable' is set, XIM is never active. */
5863     if (p_imdisable)
5864 	active = FALSE;
5865     else if (input_style & XIMPreeditPosition)
5866 	/* There is a problem in switching XIM off when preediting is used,
5867 	 * and it is not clear how this can be solved.  For now, keep XIM on
5868 	 * all the time, like it was done in Vim 5.8. */
5869 	active = TRUE;
5870 
5871 #  if defined(FEAT_EVAL)
5872     if (USE_IMACTIVATEFUNC)
5873     {
5874 	if (active != im_get_status())
5875 	{
5876 	    call_imactivatefunc(active);
5877 	    xim_has_focus = active;
5878 	}
5879 	return;
5880     }
5881 #  endif
5882 
5883     if (xic == NULL)
5884 	return;
5885 
5886     /* Remember the active state, it is needed when Vim gets keyboard focus. */
5887     xim_is_active = active;
5888     xim_set_preedit();
5889 }
5890 
5891 /*
5892  * Adjust using XIM for gaining or losing keyboard focus.  Also called when
5893  * "xim_is_active" changes.
5894  */
5895     void
5896 xim_set_focus(int focus)
5897 {
5898     if (xic == NULL)
5899 	return;
5900 
5901     /*
5902      * XIM only gets focus when the Vim window has keyboard focus and XIM has
5903      * been set active for the current mode.
5904      */
5905     if (focus && xim_is_active)
5906     {
5907 	if (!xim_has_focus)
5908 	{
5909 	    xim_has_focus = TRUE;
5910 	    XSetICFocus(xic);
5911 	}
5912     }
5913     else
5914     {
5915 	if (xim_has_focus)
5916 	{
5917 	    xim_has_focus = FALSE;
5918 	    XUnsetICFocus(xic);
5919 	}
5920     }
5921 }
5922 
5923     void
5924 im_set_position(int row UNUSED, int col UNUSED)
5925 {
5926     xim_set_preedit();
5927 }
5928 
5929 /*
5930  * Set the XIM to the current cursor position.
5931  */
5932     void
5933 xim_set_preedit(void)
5934 {
5935     XVaNestedList attr_list;
5936     XRectangle spot_area;
5937     XPoint over_spot;
5938     int line_space;
5939 
5940     if (xic == NULL)
5941 	return;
5942 
5943     xim_set_focus(TRUE);
5944 
5945     if (!xim_has_focus)
5946     {
5947 	/* hide XIM cursor */
5948 	over_spot.x = 0;
5949 	over_spot.y = -100; /* arbitrary invisible position */
5950 	attr_list = (XVaNestedList) XVaCreateNestedList(0,
5951 							XNSpotLocation,
5952 							&over_spot,
5953 							NULL);
5954 	XSetICValues(xic, XNPreeditAttributes, attr_list, NULL);
5955 	XFree(attr_list);
5956 	return;
5957     }
5958 
5959     if (input_style & XIMPreeditPosition)
5960     {
5961 	if (xim_fg_color == INVALCOLOR)
5962 	{
5963 	    xim_fg_color = gui.def_norm_pixel;
5964 	    xim_bg_color = gui.def_back_pixel;
5965 	}
5966 	over_spot.x = TEXT_X(gui.col);
5967 	over_spot.y = TEXT_Y(gui.row);
5968 	spot_area.x = 0;
5969 	spot_area.y = 0;
5970 	spot_area.height = gui.char_height * Rows;
5971 	spot_area.width  = gui.char_width * Columns;
5972 	line_space = gui.char_height;
5973 	attr_list = (XVaNestedList) XVaCreateNestedList(0,
5974 					XNSpotLocation, &over_spot,
5975 					XNForeground, (Pixel) xim_fg_color,
5976 					XNBackground, (Pixel) xim_bg_color,
5977 					XNArea, &spot_area,
5978 					XNLineSpace, line_space,
5979 					NULL);
5980 	if (XSetICValues(xic, XNPreeditAttributes, attr_list, NULL))
5981 	    emsg(_("E284: Cannot set IC values"));
5982 	XFree(attr_list);
5983     }
5984 }
5985 
5986 #  if defined(FEAT_GUI_X11)
5987 static char e_xim[] = N_("E285: Failed to create input context");
5988 #  endif
5989 
5990 #  if defined(FEAT_GUI_X11) || defined(PROTO)
5991 #   if defined(XtSpecificationRelease) && XtSpecificationRelease >= 6 && !defined(SUN_SYSTEM)
5992 #    define USE_X11R6_XIM
5993 #   endif
5994 
5995 static int xim_real_init(Window x11_window, Display *x11_display);
5996 
5997 
5998 #  ifdef USE_X11R6_XIM
5999     static void
6000 xim_instantiate_cb(
6001     Display	*display,
6002     XPointer	client_data UNUSED,
6003     XPointer	call_data UNUSED)
6004 {
6005     Window	x11_window;
6006     Display	*x11_display;
6007 
6008 #   ifdef XIM_DEBUG
6009     xim_log("xim_instantiate_cb()\n");
6010 #   endif
6011 
6012     gui_get_x11_windis(&x11_window, &x11_display);
6013     if (display != x11_display)
6014 	return;
6015 
6016     xim_real_init(x11_window, x11_display);
6017     gui_set_shellsize(FALSE, FALSE, RESIZE_BOTH);
6018     if (xic != NULL)
6019 	XUnregisterIMInstantiateCallback(x11_display, NULL, NULL, NULL,
6020 					 xim_instantiate_cb, NULL);
6021 }
6022 
6023     static void
6024 xim_destroy_cb(
6025     XIM		im UNUSED,
6026     XPointer	client_data UNUSED,
6027     XPointer	call_data UNUSED)
6028 {
6029     Window	x11_window;
6030     Display	*x11_display;
6031 
6032 #   ifdef XIM_DEBUG
6033     xim_log("xim_destroy_cb()\n");
6034    #endif
6035     gui_get_x11_windis(&x11_window, &x11_display);
6036 
6037     xic = NULL;
6038     status_area_enabled = FALSE;
6039 
6040     gui_set_shellsize(FALSE, FALSE, RESIZE_BOTH);
6041 
6042     XRegisterIMInstantiateCallback(x11_display, NULL, NULL, NULL,
6043 				   xim_instantiate_cb, NULL);
6044 }
6045 #  endif
6046 
6047     void
6048 xim_init(void)
6049 {
6050     Window	x11_window;
6051     Display	*x11_display;
6052 
6053 #  ifdef XIM_DEBUG
6054     xim_log("xim_init()\n");
6055 #  endif
6056 
6057     gui_get_x11_windis(&x11_window, &x11_display);
6058 
6059     xic = NULL;
6060 
6061     if (xim_real_init(x11_window, x11_display))
6062 	return;
6063 
6064     gui_set_shellsize(FALSE, FALSE, RESIZE_BOTH);
6065 
6066 #  ifdef USE_X11R6_XIM
6067     XRegisterIMInstantiateCallback(x11_display, NULL, NULL, NULL,
6068 				   xim_instantiate_cb, NULL);
6069 #  endif
6070 }
6071 
6072     static int
6073 xim_real_init(Window x11_window, Display *x11_display)
6074 {
6075     int		i;
6076     char	*p,
6077 		*s,
6078 		*ns,
6079 		*end,
6080 		tmp[1024];
6081 #  define IMLEN_MAX 40
6082     char	buf[IMLEN_MAX + 7];
6083     XIM		xim = NULL;
6084     XIMStyles	*xim_styles;
6085     XIMStyle	this_input_style = 0;
6086     Boolean	found;
6087     XPoint	over_spot;
6088     XVaNestedList preedit_list, status_list;
6089 
6090     input_style = 0;
6091     status_area_enabled = FALSE;
6092 
6093     if (xic != NULL)
6094 	return FALSE;
6095 
6096     if (gui.rsrc_input_method != NULL && *gui.rsrc_input_method != NUL)
6097     {
6098 	strcpy(tmp, gui.rsrc_input_method);
6099 	for (ns = s = tmp; ns != NULL && *s != NUL;)
6100 	{
6101 	    s = (char *)skipwhite((char_u *)s);
6102 	    if (*s == NUL)
6103 		break;
6104 	    if ((ns = end = strchr(s, ',')) == NULL)
6105 		end = s + strlen(s);
6106 	    while (isspace(((char_u *)end)[-1]))
6107 		end--;
6108 	    *end = NUL;
6109 
6110 	    if (strlen(s) <= IMLEN_MAX)
6111 	    {
6112 		strcpy(buf, "@im=");
6113 		strcat(buf, s);
6114 		if ((p = XSetLocaleModifiers(buf)) != NULL && *p != NUL
6115 			&& (xim = XOpenIM(x11_display, NULL, NULL, NULL))
6116 								      != NULL)
6117 		    break;
6118 	    }
6119 
6120 	    s = ns + 1;
6121 	}
6122     }
6123 
6124     if (xim == NULL && (p = XSetLocaleModifiers("")) != NULL && *p != NUL)
6125 	xim = XOpenIM(x11_display, NULL, NULL, NULL);
6126 
6127     /* This is supposed to be useful to obtain characters through
6128      * XmbLookupString() without really using a XIM. */
6129     if (xim == NULL && (p = XSetLocaleModifiers("@im=none")) != NULL
6130 								 && *p != NUL)
6131 	xim = XOpenIM(x11_display, NULL, NULL, NULL);
6132 
6133     if (xim == NULL)
6134     {
6135 	/* Only give this message when verbose is set, because too many people
6136 	 * got this message when they didn't want to use a XIM. */
6137 	if (p_verbose > 0)
6138 	{
6139 	    verbose_enter();
6140 	    emsg(_("E286: Failed to open input method"));
6141 	    verbose_leave();
6142 	}
6143 	return FALSE;
6144     }
6145 
6146 #  ifdef USE_X11R6_XIM
6147     {
6148 	XIMCallback destroy_cb;
6149 
6150 	destroy_cb.callback = xim_destroy_cb;
6151 	destroy_cb.client_data = NULL;
6152 	if (XSetIMValues(xim, XNDestroyCallback, &destroy_cb, NULL))
6153 	    emsg(_("E287: Warning: Could not set destroy callback to IM"));
6154     }
6155 #  endif
6156 
6157     if (XGetIMValues(xim, XNQueryInputStyle, &xim_styles, NULL) || !xim_styles)
6158     {
6159 	emsg(_("E288: input method doesn't support any style"));
6160 	XCloseIM(xim);
6161 	return FALSE;
6162     }
6163 
6164     found = False;
6165     strcpy(tmp, gui.rsrc_preedit_type_name);
6166     for (s = tmp; s && !found; )
6167     {
6168 	while (*s && isspace((unsigned char)*s))
6169 	    s++;
6170 	if (!*s)
6171 	    break;
6172 	if ((ns = end = strchr(s, ',')) != 0)
6173 	    ns++;
6174 	else
6175 	    end = s + strlen(s);
6176 	while (isspace((unsigned char)*end))
6177 	    end--;
6178 	*end = '\0';
6179 
6180 	if (!strcmp(s, "OverTheSpot"))
6181 	    this_input_style = (XIMPreeditPosition | XIMStatusArea);
6182 	else if (!strcmp(s, "OffTheSpot"))
6183 	    this_input_style = (XIMPreeditArea | XIMStatusArea);
6184 	else if (!strcmp(s, "Root"))
6185 	    this_input_style = (XIMPreeditNothing | XIMStatusNothing);
6186 
6187 	for (i = 0; (unsigned short)i < xim_styles->count_styles; i++)
6188 	{
6189 	    if (this_input_style == xim_styles->supported_styles[i])
6190 	    {
6191 		found = True;
6192 		break;
6193 	    }
6194 	}
6195 	if (!found)
6196 	    for (i = 0; (unsigned short)i < xim_styles->count_styles; i++)
6197 	    {
6198 		if ((xim_styles->supported_styles[i] & this_input_style)
6199 			== (this_input_style & ~XIMStatusArea))
6200 		{
6201 		    this_input_style &= ~XIMStatusArea;
6202 		    found = True;
6203 		    break;
6204 		}
6205 	    }
6206 
6207 	s = ns;
6208     }
6209     XFree(xim_styles);
6210 
6211     if (!found)
6212     {
6213 	/* Only give this message when verbose is set, because too many people
6214 	 * got this message when they didn't want to use a XIM. */
6215 	if (p_verbose > 0)
6216 	{
6217 	    verbose_enter();
6218 	    emsg(_("E289: input method doesn't support my preedit type"));
6219 	    verbose_leave();
6220 	}
6221 	XCloseIM(xim);
6222 	return FALSE;
6223     }
6224 
6225     over_spot.x = TEXT_X(gui.col);
6226     over_spot.y = TEXT_Y(gui.row);
6227     input_style = this_input_style;
6228 
6229     /* A crash was reported when trying to pass gui.norm_font as XNFontSet,
6230      * thus that has been removed.  Hopefully the default works... */
6231 #  ifdef FEAT_XFONTSET
6232     if (gui.fontset != NOFONTSET)
6233     {
6234 	preedit_list = XVaCreateNestedList(0,
6235 				XNSpotLocation, &over_spot,
6236 				XNForeground, (Pixel)gui.def_norm_pixel,
6237 				XNBackground, (Pixel)gui.def_back_pixel,
6238 				XNFontSet, (XFontSet)gui.fontset,
6239 				NULL);
6240 	status_list = XVaCreateNestedList(0,
6241 				XNForeground, (Pixel)gui.def_norm_pixel,
6242 				XNBackground, (Pixel)gui.def_back_pixel,
6243 				XNFontSet, (XFontSet)gui.fontset,
6244 				NULL);
6245     }
6246     else
6247 #  endif
6248     {
6249 	preedit_list = XVaCreateNestedList(0,
6250 				XNSpotLocation, &over_spot,
6251 				XNForeground, (Pixel)gui.def_norm_pixel,
6252 				XNBackground, (Pixel)gui.def_back_pixel,
6253 				NULL);
6254 	status_list = XVaCreateNestedList(0,
6255 				XNForeground, (Pixel)gui.def_norm_pixel,
6256 				XNBackground, (Pixel)gui.def_back_pixel,
6257 				NULL);
6258     }
6259 
6260     xic = XCreateIC(xim,
6261 		    XNInputStyle, input_style,
6262 		    XNClientWindow, x11_window,
6263 		    XNFocusWindow, gui.wid,
6264 		    XNPreeditAttributes, preedit_list,
6265 		    XNStatusAttributes, status_list,
6266 		    NULL);
6267     XFree(status_list);
6268     XFree(preedit_list);
6269     if (xic != NULL)
6270     {
6271 	if (input_style & XIMStatusArea)
6272 	{
6273 	    xim_set_status_area();
6274 	    status_area_enabled = TRUE;
6275 	}
6276 	else
6277 	    gui_set_shellsize(FALSE, FALSE, RESIZE_BOTH);
6278     }
6279     else
6280     {
6281 	if (!is_not_a_term())
6282 	    emsg(_(e_xim));
6283 	XCloseIM(xim);
6284 	return FALSE;
6285     }
6286 
6287     return TRUE;
6288 }
6289 
6290 #  endif /* FEAT_GUI_X11 */
6291 
6292 /*
6293  * Get IM status.  When IM is on, return TRUE.  Else return FALSE.
6294  * FIXME: This doesn't work correctly: Having focus doesn't always mean XIM is
6295  * active, when not having focus XIM may still be active (e.g., when using a
6296  * tear-off menu item).
6297  */
6298     int
6299 im_get_status(void)
6300 {
6301 #  ifdef FEAT_EVAL
6302     if (USE_IMSTATUSFUNC)
6303 	return call_imstatusfunc();
6304 #  endif
6305     return xim_has_focus;
6306 }
6307 
6308 # endif /* !FEAT_GUI_GTK */
6309 
6310 # if !defined(FEAT_GUI_GTK) || defined(PROTO)
6311 /*
6312  * Set up the status area.
6313  *
6314  * This should use a separate Widget, but that seems not possible, because
6315  * preedit_area and status_area should be set to the same window as for the
6316  * text input.  Unfortunately this means the status area pollutes the text
6317  * window...
6318  */
6319     void
6320 xim_set_status_area(void)
6321 {
6322     XVaNestedList preedit_list = 0, status_list = 0, list = 0;
6323     XRectangle pre_area, status_area;
6324 
6325     if (xic == NULL)
6326 	return;
6327 
6328     if (input_style & XIMStatusArea)
6329     {
6330 	if (input_style & XIMPreeditArea)
6331 	{
6332 	    XRectangle *needed_rect;
6333 
6334 	    /* to get status_area width */
6335 	    status_list = XVaCreateNestedList(0, XNAreaNeeded,
6336 					      &needed_rect, NULL);
6337 	    XGetICValues(xic, XNStatusAttributes, status_list, NULL);
6338 	    XFree(status_list);
6339 
6340 	    status_area.width = needed_rect->width;
6341 	}
6342 	else
6343 	    status_area.width = gui.char_width * Columns;
6344 
6345 	status_area.x = 0;
6346 	status_area.y = gui.char_height * Rows + gui.border_offset;
6347 	if (gui.which_scrollbars[SBAR_BOTTOM])
6348 	    status_area.y += gui.scrollbar_height;
6349 #ifdef FEAT_MENU
6350 	if (gui.menu_is_active)
6351 	    status_area.y += gui.menu_height;
6352 #endif
6353 	status_area.height = gui.char_height;
6354 	status_list = XVaCreateNestedList(0, XNArea, &status_area, NULL);
6355     }
6356     else
6357     {
6358 	status_area.x = 0;
6359 	status_area.y = gui.char_height * Rows + gui.border_offset;
6360 	if (gui.which_scrollbars[SBAR_BOTTOM])
6361 	    status_area.y += gui.scrollbar_height;
6362 #ifdef FEAT_MENU
6363 	if (gui.menu_is_active)
6364 	    status_area.y += gui.menu_height;
6365 #endif
6366 	status_area.width = 0;
6367 	status_area.height = gui.char_height;
6368     }
6369 
6370     if (input_style & XIMPreeditArea)   /* off-the-spot */
6371     {
6372 	pre_area.x = status_area.x + status_area.width;
6373 	pre_area.y = gui.char_height * Rows + gui.border_offset;
6374 	pre_area.width = gui.char_width * Columns - pre_area.x;
6375 	if (gui.which_scrollbars[SBAR_BOTTOM])
6376 	    pre_area.y += gui.scrollbar_height;
6377 #ifdef FEAT_MENU
6378 	if (gui.menu_is_active)
6379 	    pre_area.y += gui.menu_height;
6380 #endif
6381 	pre_area.height = gui.char_height;
6382 	preedit_list = XVaCreateNestedList(0, XNArea, &pre_area, NULL);
6383     }
6384     else if (input_style & XIMPreeditPosition)   /* over-the-spot */
6385     {
6386 	pre_area.x = 0;
6387 	pre_area.y = 0;
6388 	pre_area.height = gui.char_height * Rows;
6389 	pre_area.width = gui.char_width * Columns;
6390 	preedit_list = XVaCreateNestedList(0, XNArea, &pre_area, NULL);
6391     }
6392 
6393     if (preedit_list && status_list)
6394 	list = XVaCreateNestedList(0, XNPreeditAttributes, preedit_list,
6395 				   XNStatusAttributes, status_list, NULL);
6396     else if (preedit_list)
6397 	list = XVaCreateNestedList(0, XNPreeditAttributes, preedit_list,
6398 				   NULL);
6399     else if (status_list)
6400 	list = XVaCreateNestedList(0, XNStatusAttributes, status_list,
6401 				   NULL);
6402     else
6403 	list = NULL;
6404 
6405     if (list)
6406     {
6407 	XSetICValues(xic, XNVaNestedList, list, NULL);
6408 	XFree(list);
6409     }
6410     if (status_list)
6411 	XFree(status_list);
6412     if (preedit_list)
6413 	XFree(preedit_list);
6414 }
6415 
6416     int
6417 xim_get_status_area_height(void)
6418 {
6419     if (status_area_enabled)
6420 	return gui.char_height;
6421     return 0;
6422 }
6423 # endif
6424 
6425 #else /* !defined(FEAT_XIM) */
6426 
6427 # ifdef IME_WITHOUT_XIM
6428 static int im_was_set_active = FALSE;
6429 
6430     int
6431 im_get_status(void)
6432 {
6433 #  if defined(FEAT_EVAL)
6434     if (USE_IMSTATUSFUNC)
6435 	return call_imstatusfunc();
6436 #  endif
6437     return im_was_set_active;
6438 }
6439 
6440     void
6441 im_set_active(int active_arg)
6442 {
6443 #  if defined(FEAT_EVAL)
6444     int	    active = !p_imdisable && active_arg;
6445 
6446     if (USE_IMACTIVATEFUNC && active != im_get_status())
6447     {
6448 	call_imactivatefunc(active);
6449 	im_was_set_active = active;
6450     }
6451 #  endif
6452 }
6453 
6454 #  ifdef FEAT_GUI
6455     void
6456 im_set_position(int row UNUSED, int col UNUSED)
6457 {
6458 }
6459 #  endif
6460 # endif
6461 
6462 #endif /* FEAT_XIM */
6463 
6464 
6465 /*
6466  * Setup "vcp" for conversion from "from" to "to".
6467  * The names must have been made canonical with enc_canonize().
6468  * vcp->vc_type must have been initialized to CONV_NONE.
6469  * Note: cannot be used for conversion from/to ucs-2 and ucs-4 (will use utf-8
6470  * instead).
6471  * Afterwards invoke with "from" and "to" equal to NULL to cleanup.
6472  * Return FAIL when conversion is not supported, OK otherwise.
6473  */
6474     int
6475 convert_setup(vimconv_T *vcp, char_u *from, char_u *to)
6476 {
6477     return convert_setup_ext(vcp, from, TRUE, to, TRUE);
6478 }
6479 
6480 /*
6481  * As convert_setup(), but only when from_unicode_is_utf8 is TRUE will all
6482  * "from" unicode charsets be considered utf-8.  Same for "to".
6483  */
6484     int
6485 convert_setup_ext(
6486     vimconv_T	*vcp,
6487     char_u	*from,
6488     int		from_unicode_is_utf8,
6489     char_u	*to,
6490     int		to_unicode_is_utf8)
6491 {
6492     int		from_prop;
6493     int		to_prop;
6494     int		from_is_utf8;
6495     int		to_is_utf8;
6496 
6497     /* Reset to no conversion. */
6498 #ifdef USE_ICONV
6499     if (vcp->vc_type == CONV_ICONV && vcp->vc_fd != (iconv_t)-1)
6500 	iconv_close(vcp->vc_fd);
6501 #endif
6502     vcp->vc_type = CONV_NONE;
6503     vcp->vc_factor = 1;
6504     vcp->vc_fail = FALSE;
6505 
6506     /* No conversion when one of the names is empty or they are equal. */
6507     if (from == NULL || *from == NUL || to == NULL || *to == NUL
6508 						     || STRCMP(from, to) == 0)
6509 	return OK;
6510 
6511     from_prop = enc_canon_props(from);
6512     to_prop = enc_canon_props(to);
6513     if (from_unicode_is_utf8)
6514 	from_is_utf8 = from_prop & ENC_UNICODE;
6515     else
6516 	from_is_utf8 = from_prop == ENC_UNICODE;
6517     if (to_unicode_is_utf8)
6518 	to_is_utf8 = to_prop & ENC_UNICODE;
6519     else
6520 	to_is_utf8 = to_prop == ENC_UNICODE;
6521 
6522     if ((from_prop & ENC_LATIN1) && to_is_utf8)
6523     {
6524 	/* Internal latin1 -> utf-8 conversion. */
6525 	vcp->vc_type = CONV_TO_UTF8;
6526 	vcp->vc_factor = 2;	/* up to twice as long */
6527     }
6528     else if ((from_prop & ENC_LATIN9) && to_is_utf8)
6529     {
6530 	/* Internal latin9 -> utf-8 conversion. */
6531 	vcp->vc_type = CONV_9_TO_UTF8;
6532 	vcp->vc_factor = 3;	/* up to three as long (euro sign) */
6533     }
6534     else if (from_is_utf8 && (to_prop & ENC_LATIN1))
6535     {
6536 	/* Internal utf-8 -> latin1 conversion. */
6537 	vcp->vc_type = CONV_TO_LATIN1;
6538     }
6539     else if (from_is_utf8 && (to_prop & ENC_LATIN9))
6540     {
6541 	/* Internal utf-8 -> latin9 conversion. */
6542 	vcp->vc_type = CONV_TO_LATIN9;
6543     }
6544 #ifdef MSWIN
6545     /* Win32-specific codepage <-> codepage conversion without iconv. */
6546     else if ((from_is_utf8 || encname2codepage(from) > 0)
6547 	    && (to_is_utf8 || encname2codepage(to) > 0))
6548     {
6549 	vcp->vc_type = CONV_CODEPAGE;
6550 	vcp->vc_factor = 2;	/* up to twice as long */
6551 	vcp->vc_cpfrom = from_is_utf8 ? 0 : encname2codepage(from);
6552 	vcp->vc_cpto = to_is_utf8 ? 0 : encname2codepage(to);
6553     }
6554 #endif
6555 #ifdef MACOS_CONVERT
6556     else if ((from_prop & ENC_MACROMAN) && (to_prop & ENC_LATIN1))
6557     {
6558 	vcp->vc_type = CONV_MAC_LATIN1;
6559     }
6560     else if ((from_prop & ENC_MACROMAN) && to_is_utf8)
6561     {
6562 	vcp->vc_type = CONV_MAC_UTF8;
6563 	vcp->vc_factor = 2;	/* up to twice as long */
6564     }
6565     else if ((from_prop & ENC_LATIN1) && (to_prop & ENC_MACROMAN))
6566     {
6567 	vcp->vc_type = CONV_LATIN1_MAC;
6568     }
6569     else if (from_is_utf8 && (to_prop & ENC_MACROMAN))
6570     {
6571 	vcp->vc_type = CONV_UTF8_MAC;
6572     }
6573 #endif
6574 #ifdef USE_ICONV
6575     else
6576     {
6577 	/* Use iconv() for conversion. */
6578 	vcp->vc_fd = (iconv_t)my_iconv_open(
6579 		to_is_utf8 ? (char_u *)"utf-8" : to,
6580 		from_is_utf8 ? (char_u *)"utf-8" : from);
6581 	if (vcp->vc_fd != (iconv_t)-1)
6582 	{
6583 	    vcp->vc_type = CONV_ICONV;
6584 	    vcp->vc_factor = 4;	/* could be longer too... */
6585 	}
6586     }
6587 #endif
6588     if (vcp->vc_type == CONV_NONE)
6589 	return FAIL;
6590 
6591     return OK;
6592 }
6593 
6594 #if defined(FEAT_GUI) || defined(AMIGA) || defined(MSWIN) \
6595 	|| defined(PROTO)
6596 /*
6597  * Do conversion on typed input characters in-place.
6598  * The input and output are not NUL terminated!
6599  * Returns the length after conversion.
6600  */
6601     int
6602 convert_input(char_u *ptr, int len, int maxlen)
6603 {
6604     return convert_input_safe(ptr, len, maxlen, NULL, NULL);
6605 }
6606 #endif
6607 
6608 /*
6609  * Like convert_input(), but when there is an incomplete byte sequence at the
6610  * end return that as an allocated string in "restp" and set "*restlenp" to
6611  * the length.  If "restp" is NULL it is not used.
6612  */
6613     int
6614 convert_input_safe(
6615     char_u	*ptr,
6616     int		len,
6617     int		maxlen,
6618     char_u	**restp,
6619     int		*restlenp)
6620 {
6621     char_u	*d;
6622     int		dlen = len;
6623     int		unconvertlen = 0;
6624 
6625     d = string_convert_ext(&input_conv, ptr, &dlen,
6626 					restp == NULL ? NULL : &unconvertlen);
6627     if (d != NULL)
6628     {
6629 	if (dlen <= maxlen)
6630 	{
6631 	    if (unconvertlen > 0)
6632 	    {
6633 		/* Move the unconverted characters to allocated memory. */
6634 		*restp = alloc(unconvertlen);
6635 		if (*restp != NULL)
6636 		    mch_memmove(*restp, ptr + len - unconvertlen, unconvertlen);
6637 		*restlenp = unconvertlen;
6638 	    }
6639 	    mch_memmove(ptr, d, dlen);
6640 	}
6641 	else
6642 	    /* result is too long, keep the unconverted text (the caller must
6643 	     * have done something wrong!) */
6644 	    dlen = len;
6645 	vim_free(d);
6646     }
6647     return dlen;
6648 }
6649 
6650 /*
6651  * Convert text "ptr[*lenp]" according to "vcp".
6652  * Returns the result in allocated memory and sets "*lenp".
6653  * When "lenp" is NULL, use NUL terminated strings.
6654  * Illegal chars are often changed to "?", unless vcp->vc_fail is set.
6655  * When something goes wrong, NULL is returned and "*lenp" is unchanged.
6656  */
6657     char_u *
6658 string_convert(
6659     vimconv_T	*vcp,
6660     char_u	*ptr,
6661     int		*lenp)
6662 {
6663     return string_convert_ext(vcp, ptr, lenp, NULL);
6664 }
6665 
6666 /*
6667  * Like string_convert(), but when "unconvlenp" is not NULL and there are is
6668  * an incomplete sequence at the end it is not converted and "*unconvlenp" is
6669  * set to the number of remaining bytes.
6670  */
6671     char_u *
6672 string_convert_ext(
6673     vimconv_T	*vcp,
6674     char_u	*ptr,
6675     int		*lenp,
6676     int		*unconvlenp)
6677 {
6678     char_u	*retval = NULL;
6679     char_u	*d;
6680     int		len;
6681     int		i;
6682     int		l;
6683     int		c;
6684 
6685     if (lenp == NULL)
6686 	len = (int)STRLEN(ptr);
6687     else
6688 	len = *lenp;
6689     if (len == 0)
6690 	return vim_strsave((char_u *)"");
6691 
6692     switch (vcp->vc_type)
6693     {
6694 	case CONV_TO_UTF8:	/* latin1 to utf-8 conversion */
6695 	    retval = alloc(len * 2 + 1);
6696 	    if (retval == NULL)
6697 		break;
6698 	    d = retval;
6699 	    for (i = 0; i < len; ++i)
6700 	    {
6701 		c = ptr[i];
6702 		if (c < 0x80)
6703 		    *d++ = c;
6704 		else
6705 		{
6706 		    *d++ = 0xc0 + ((unsigned)c >> 6);
6707 		    *d++ = 0x80 + (c & 0x3f);
6708 		}
6709 	    }
6710 	    *d = NUL;
6711 	    if (lenp != NULL)
6712 		*lenp = (int)(d - retval);
6713 	    break;
6714 
6715 	case CONV_9_TO_UTF8:	/* latin9 to utf-8 conversion */
6716 	    retval = alloc(len * 3 + 1);
6717 	    if (retval == NULL)
6718 		break;
6719 	    d = retval;
6720 	    for (i = 0; i < len; ++i)
6721 	    {
6722 		c = ptr[i];
6723 		switch (c)
6724 		{
6725 		    case 0xa4: c = 0x20ac; break;   /* euro */
6726 		    case 0xa6: c = 0x0160; break;   /* S hat */
6727 		    case 0xa8: c = 0x0161; break;   /* S -hat */
6728 		    case 0xb4: c = 0x017d; break;   /* Z hat */
6729 		    case 0xb8: c = 0x017e; break;   /* Z -hat */
6730 		    case 0xbc: c = 0x0152; break;   /* OE */
6731 		    case 0xbd: c = 0x0153; break;   /* oe */
6732 		    case 0xbe: c = 0x0178; break;   /* Y */
6733 		}
6734 		d += utf_char2bytes(c, d);
6735 	    }
6736 	    *d = NUL;
6737 	    if (lenp != NULL)
6738 		*lenp = (int)(d - retval);
6739 	    break;
6740 
6741 	case CONV_TO_LATIN1:	/* utf-8 to latin1 conversion */
6742 	case CONV_TO_LATIN9:	/* utf-8 to latin9 conversion */
6743 	    retval = alloc(len + 1);
6744 	    if (retval == NULL)
6745 		break;
6746 	    d = retval;
6747 	    for (i = 0; i < len; ++i)
6748 	    {
6749 		l = utf_ptr2len_len(ptr + i, len - i);
6750 		if (l == 0)
6751 		    *d++ = NUL;
6752 		else if (l == 1)
6753 		{
6754 		    int l_w = utf8len_tab_zero[ptr[i]];
6755 
6756 		    if (l_w == 0)
6757 		    {
6758 			/* Illegal utf-8 byte cannot be converted */
6759 			vim_free(retval);
6760 			return NULL;
6761 		    }
6762 		    if (unconvlenp != NULL && l_w > len - i)
6763 		    {
6764 			/* Incomplete sequence at the end. */
6765 			*unconvlenp = len - i;
6766 			break;
6767 		    }
6768 		    *d++ = ptr[i];
6769 		}
6770 		else
6771 		{
6772 		    c = utf_ptr2char(ptr + i);
6773 		    if (vcp->vc_type == CONV_TO_LATIN9)
6774 			switch (c)
6775 			{
6776 			    case 0x20ac: c = 0xa4; break;   /* euro */
6777 			    case 0x0160: c = 0xa6; break;   /* S hat */
6778 			    case 0x0161: c = 0xa8; break;   /* S -hat */
6779 			    case 0x017d: c = 0xb4; break;   /* Z hat */
6780 			    case 0x017e: c = 0xb8; break;   /* Z -hat */
6781 			    case 0x0152: c = 0xbc; break;   /* OE */
6782 			    case 0x0153: c = 0xbd; break;   /* oe */
6783 			    case 0x0178: c = 0xbe; break;   /* Y */
6784 			    case 0xa4:
6785 			    case 0xa6:
6786 			    case 0xa8:
6787 			    case 0xb4:
6788 			    case 0xb8:
6789 			    case 0xbc:
6790 			    case 0xbd:
6791 			    case 0xbe: c = 0x100; break; /* not in latin9 */
6792 			}
6793 		    if (!utf_iscomposing(c))	/* skip composing chars */
6794 		    {
6795 			if (c < 0x100)
6796 			    *d++ = c;
6797 			else if (vcp->vc_fail)
6798 			{
6799 			    vim_free(retval);
6800 			    return NULL;
6801 			}
6802 			else
6803 			{
6804 			    *d++ = 0xbf;
6805 			    if (utf_char2cells(c) > 1)
6806 				*d++ = '?';
6807 			}
6808 		    }
6809 		    i += l - 1;
6810 		}
6811 	    }
6812 	    *d = NUL;
6813 	    if (lenp != NULL)
6814 		*lenp = (int)(d - retval);
6815 	    break;
6816 
6817 # ifdef MACOS_CONVERT
6818 	case CONV_MAC_LATIN1:
6819 	    retval = mac_string_convert(ptr, len, lenp, vcp->vc_fail,
6820 					'm', 'l', unconvlenp);
6821 	    break;
6822 
6823 	case CONV_LATIN1_MAC:
6824 	    retval = mac_string_convert(ptr, len, lenp, vcp->vc_fail,
6825 					'l', 'm', unconvlenp);
6826 	    break;
6827 
6828 	case CONV_MAC_UTF8:
6829 	    retval = mac_string_convert(ptr, len, lenp, vcp->vc_fail,
6830 					'm', 'u', unconvlenp);
6831 	    break;
6832 
6833 	case CONV_UTF8_MAC:
6834 	    retval = mac_string_convert(ptr, len, lenp, vcp->vc_fail,
6835 					'u', 'm', unconvlenp);
6836 	    break;
6837 # endif
6838 
6839 # ifdef USE_ICONV
6840 	case CONV_ICONV:	/* conversion with output_conv.vc_fd */
6841 	    retval = iconv_string(vcp, ptr, len, unconvlenp, lenp);
6842 	    break;
6843 # endif
6844 # ifdef MSWIN
6845 	case CONV_CODEPAGE:		/* codepage -> codepage */
6846 	{
6847 	    int		retlen;
6848 	    int		tmp_len;
6849 	    short_u	*tmp;
6850 
6851 	    /* 1. codepage/UTF-8  ->  ucs-2. */
6852 	    if (vcp->vc_cpfrom == 0)
6853 		tmp_len = utf8_to_utf16(ptr, len, NULL, NULL);
6854 	    else
6855 	    {
6856 		tmp_len = MultiByteToWideChar(vcp->vc_cpfrom,
6857 					unconvlenp ? MB_ERR_INVALID_CHARS : 0,
6858 					(char *)ptr, len, 0, 0);
6859 		if (tmp_len == 0
6860 			&& GetLastError() == ERROR_NO_UNICODE_TRANSLATION)
6861 		{
6862 		    if (lenp != NULL)
6863 			*lenp = 0;
6864 		    if (unconvlenp != NULL)
6865 			*unconvlenp = len;
6866 		    retval = alloc(1);
6867 		    if (retval)
6868 			retval[0] = NUL;
6869 		    return retval;
6870 		}
6871 	    }
6872 	    tmp = (short_u *)alloc(sizeof(short_u) * tmp_len);
6873 	    if (tmp == NULL)
6874 		break;
6875 	    if (vcp->vc_cpfrom == 0)
6876 		utf8_to_utf16(ptr, len, tmp, unconvlenp);
6877 	    else
6878 		MultiByteToWideChar(vcp->vc_cpfrom, 0,
6879 			(char *)ptr, len, tmp, tmp_len);
6880 
6881 	    /* 2. ucs-2  ->  codepage/UTF-8. */
6882 	    if (vcp->vc_cpto == 0)
6883 		retlen = utf16_to_utf8(tmp, tmp_len, NULL);
6884 	    else
6885 		retlen = WideCharToMultiByte(vcp->vc_cpto, 0,
6886 						    tmp, tmp_len, 0, 0, 0, 0);
6887 	    retval = alloc(retlen + 1);
6888 	    if (retval != NULL)
6889 	    {
6890 		if (vcp->vc_cpto == 0)
6891 		    utf16_to_utf8(tmp, tmp_len, retval);
6892 		else
6893 		    WideCharToMultiByte(vcp->vc_cpto, 0,
6894 					  tmp, tmp_len,
6895 					  (char *)retval, retlen, 0, 0);
6896 		retval[retlen] = NUL;
6897 		if (lenp != NULL)
6898 		    *lenp = retlen;
6899 	    }
6900 	    vim_free(tmp);
6901 	    break;
6902 	}
6903 # endif
6904     }
6905 
6906     return retval;
6907 }
6908