xref: /vim-8.2.3635/src/mbyte.c (revision eae1b91f)
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, 0x16fe3},
1478 	{0x17000, 0x187f7},
1479 	{0x18800, 0x18af2},
1480 	{0x1b000, 0x1b11e},
1481 	{0x1b150, 0x1b152},
1482 	{0x1b164, 0x1b167},
1483 	{0x1b170, 0x1b2fb},
1484 	{0x1f004, 0x1f004},
1485 	{0x1f0cf, 0x1f0cf},
1486 	{0x1f18e, 0x1f18e},
1487 	{0x1f191, 0x1f19a},
1488 	{0x1f200, 0x1f202},
1489 	{0x1f210, 0x1f23b},
1490 	{0x1f240, 0x1f248},
1491 	{0x1f250, 0x1f251},
1492 	{0x1f260, 0x1f265},
1493 	{0x1f300, 0x1f320},
1494 	{0x1f32d, 0x1f335},
1495 	{0x1f337, 0x1f37c},
1496 	{0x1f37e, 0x1f393},
1497 	{0x1f3a0, 0x1f3ca},
1498 	{0x1f3cf, 0x1f3d3},
1499 	{0x1f3e0, 0x1f3f0},
1500 	{0x1f3f4, 0x1f3f4},
1501 	{0x1f3f8, 0x1f43e},
1502 	{0x1f440, 0x1f440},
1503 	{0x1f442, 0x1f4fc},
1504 	{0x1f4ff, 0x1f53d},
1505 	{0x1f54b, 0x1f54e},
1506 	{0x1f550, 0x1f567},
1507 	{0x1f57a, 0x1f57a},
1508 	{0x1f595, 0x1f596},
1509 	{0x1f5a4, 0x1f5a4},
1510 	{0x1f5fb, 0x1f64f},
1511 	{0x1f680, 0x1f6c5},
1512 	{0x1f6cc, 0x1f6cc},
1513 	{0x1f6d0, 0x1f6d2},
1514 	{0x1f6d5, 0x1f6d5},
1515 	{0x1f6eb, 0x1f6ec},
1516 	{0x1f6f4, 0x1f6fa},
1517 	{0x1f7e0, 0x1f7eb},
1518 	{0x1f90d, 0x1f971},
1519 	{0x1f973, 0x1f976},
1520 	{0x1f97a, 0x1f9a2},
1521 	{0x1f9a5, 0x1f9aa},
1522 	{0x1f9ae, 0x1f9ca},
1523 	{0x1f9cd, 0x1f9ff},
1524 	{0x1fa70, 0x1fa73},
1525 	{0x1fa78, 0x1fa7a},
1526 	{0x1fa80, 0x1fa82},
1527 	{0x1fa90, 0x1fa95},
1528 	{0x20000, 0x2fffd},
1529 	{0x30000, 0x3fffd}
1530     };
1531 
1532     /* Sorted list of non-overlapping intervals of Emoji characters that don't
1533      * have ambiguous or double width,
1534      * based on http://unicode.org/emoji/charts/emoji-list.html */
1535     static struct interval emoji_width[] =
1536     {
1537 	{0x1f1e6, 0x1f1ff},
1538 	{0x1f321, 0x1f321},
1539 	{0x1f324, 0x1f32c},
1540 	{0x1f336, 0x1f336},
1541 	{0x1f37d, 0x1f37d},
1542 	{0x1f396, 0x1f397},
1543 	{0x1f399, 0x1f39b},
1544 	{0x1f39e, 0x1f39f},
1545 	{0x1f3cb, 0x1f3ce},
1546 	{0x1f3d4, 0x1f3df},
1547 	{0x1f3f3, 0x1f3f5},
1548 	{0x1f3f7, 0x1f3f7},
1549 	{0x1f43f, 0x1f43f},
1550 	{0x1f441, 0x1f441},
1551 	{0x1f4fd, 0x1f4fd},
1552 	{0x1f549, 0x1f54a},
1553 	{0x1f56f, 0x1f570},
1554 	{0x1f573, 0x1f579},
1555 	{0x1f587, 0x1f587},
1556 	{0x1f58a, 0x1f58d},
1557 	{0x1f590, 0x1f590},
1558 	{0x1f5a5, 0x1f5a5},
1559 	{0x1f5a8, 0x1f5a8},
1560 	{0x1f5b1, 0x1f5b2},
1561 	{0x1f5bc, 0x1f5bc},
1562 	{0x1f5c2, 0x1f5c4},
1563 	{0x1f5d1, 0x1f5d3},
1564 	{0x1f5dc, 0x1f5de},
1565 	{0x1f5e1, 0x1f5e1},
1566 	{0x1f5e3, 0x1f5e3},
1567 	{0x1f5e8, 0x1f5e8},
1568 	{0x1f5ef, 0x1f5ef},
1569 	{0x1f5f3, 0x1f5f3},
1570 	{0x1f5fa, 0x1f5fa},
1571 	{0x1f6cb, 0x1f6cf},
1572 	{0x1f6e0, 0x1f6e5},
1573 	{0x1f6e9, 0x1f6e9},
1574 	{0x1f6f0, 0x1f6f0},
1575 	{0x1f6f3, 0x1f6f3}
1576     };
1577 
1578     if (c >= 0x100)
1579     {
1580 #ifdef USE_WCHAR_FUNCTIONS
1581 	/*
1582 	 * Assume the library function wcwidth() works better than our own
1583 	 * stuff.  It should return 1 for ambiguous width chars!
1584 	 */
1585 	int	n = wcwidth(c);
1586 
1587 	if (n < 0)
1588 	    return 6;		/* unprintable, displays <xxxx> */
1589 	if (n > 1)
1590 	    return n;
1591 #else
1592 	if (!utf_printable(c))
1593 	    return 6;		/* unprintable, displays <xxxx> */
1594 	if (intable(doublewidth, sizeof(doublewidth), c))
1595 	    return 2;
1596 #endif
1597 	if (p_emoji && intable(emoji_width, sizeof(emoji_width), c))
1598 	    return 2;
1599     }
1600 
1601     /* Characters below 0x100 are influenced by 'isprint' option */
1602     else if (c >= 0x80 && !vim_isprintc(c))
1603 	return 4;		/* unprintable, displays <xx> */
1604 
1605     if (c >= 0x80 && *p_ambw == 'd' && intable(ambiguous, sizeof(ambiguous), c))
1606 	return 2;
1607 
1608     return 1;
1609 }
1610 
1611 /*
1612  * mb_ptr2cells() function pointer.
1613  * Return the number of display cells character at "*p" occupies.
1614  * This doesn't take care of unprintable characters, use ptr2cells() for that.
1615  */
1616     int
1617 latin_ptr2cells(char_u *p UNUSED)
1618 {
1619     return 1;
1620 }
1621 
1622     int
1623 utf_ptr2cells(
1624     char_u	*p)
1625 {
1626     int		c;
1627 
1628     /* Need to convert to a wide character. */
1629     if (*p >= 0x80)
1630     {
1631 	c = utf_ptr2char(p);
1632 	/* An illegal byte is displayed as <xx>. */
1633 	if (utf_ptr2len(p) == 1 || c == NUL)
1634 	    return 4;
1635 	/* If the char is ASCII it must be an overlong sequence. */
1636 	if (c < 0x80)
1637 	    return char2cells(c);
1638 	return utf_char2cells(c);
1639     }
1640     return 1;
1641 }
1642 
1643     int
1644 dbcs_ptr2cells(char_u *p)
1645 {
1646     /* Number of cells is equal to number of bytes, except for euc-jp when
1647      * the first byte is 0x8e. */
1648     if (enc_dbcs == DBCS_JPNU && *p == 0x8e)
1649 	return 1;
1650     return MB_BYTE2LEN(*p);
1651 }
1652 
1653 /*
1654  * mb_ptr2cells_len() function pointer.
1655  * Like mb_ptr2cells(), but limit string length to "size".
1656  * For an empty string or truncated character returns 1.
1657  */
1658     int
1659 latin_ptr2cells_len(char_u *p UNUSED, int size UNUSED)
1660 {
1661     return 1;
1662 }
1663 
1664     static int
1665 utf_ptr2cells_len(char_u *p, int size)
1666 {
1667     int		c;
1668 
1669     /* Need to convert to a wide character. */
1670     if (size > 0 && *p >= 0x80)
1671     {
1672 	if (utf_ptr2len_len(p, size) < utf8len_tab[*p])
1673 	    return 1;  /* truncated */
1674 	c = utf_ptr2char(p);
1675 	/* An illegal byte is displayed as <xx>. */
1676 	if (utf_ptr2len(p) == 1 || c == NUL)
1677 	    return 4;
1678 	/* If the char is ASCII it must be an overlong sequence. */
1679 	if (c < 0x80)
1680 	    return char2cells(c);
1681 	return utf_char2cells(c);
1682     }
1683     return 1;
1684 }
1685 
1686     static int
1687 dbcs_ptr2cells_len(char_u *p, int size)
1688 {
1689     /* Number of cells is equal to number of bytes, except for euc-jp when
1690      * the first byte is 0x8e. */
1691     if (size <= 1 || (enc_dbcs == DBCS_JPNU && *p == 0x8e))
1692 	return 1;
1693     return MB_BYTE2LEN(*p);
1694 }
1695 
1696 /*
1697  * mb_char2cells() function pointer.
1698  * Return the number of display cells character "c" occupies.
1699  * Only takes care of multi-byte chars, not "^C" and such.
1700  */
1701     int
1702 latin_char2cells(int c UNUSED)
1703 {
1704     return 1;
1705 }
1706 
1707     static int
1708 dbcs_char2cells(int c)
1709 {
1710     /* Number of cells is equal to number of bytes, except for euc-jp when
1711      * the first byte is 0x8e. */
1712     if (enc_dbcs == DBCS_JPNU && ((unsigned)c >> 8) == 0x8e)
1713 	return 1;
1714     /* use the first byte */
1715     return MB_BYTE2LEN((unsigned)c >> 8);
1716 }
1717 
1718 /*
1719  * Return the number of cells occupied by string "p".
1720  * Stop at a NUL character.  When "len" >= 0 stop at character "p[len]".
1721  */
1722     int
1723 mb_string2cells(char_u *p, int len)
1724 {
1725     int i;
1726     int clen = 0;
1727 
1728     for (i = 0; (len < 0 || i < len) && p[i] != NUL; i += (*mb_ptr2len)(p + i))
1729 	clen += (*mb_ptr2cells)(p + i);
1730     return clen;
1731 }
1732 
1733 /*
1734  * mb_off2cells() function pointer.
1735  * Return number of display cells for char at ScreenLines[off].
1736  * We make sure that the offset used is less than "max_off".
1737  */
1738     int
1739 latin_off2cells(unsigned off UNUSED, unsigned max_off UNUSED)
1740 {
1741     return 1;
1742 }
1743 
1744     int
1745 dbcs_off2cells(unsigned off, unsigned max_off)
1746 {
1747     /* never check beyond end of the line */
1748     if (off >= max_off)
1749 	return 1;
1750 
1751     /* Number of cells is equal to number of bytes, except for euc-jp when
1752      * the first byte is 0x8e. */
1753     if (enc_dbcs == DBCS_JPNU && ScreenLines[off] == 0x8e)
1754 	return 1;
1755     return MB_BYTE2LEN(ScreenLines[off]);
1756 }
1757 
1758     int
1759 utf_off2cells(unsigned off, unsigned max_off)
1760 {
1761     return (off + 1 < max_off && ScreenLines[off + 1] == 0) ? 2 : 1;
1762 }
1763 
1764 /*
1765  * mb_ptr2char() function pointer.
1766  * Convert a byte sequence into a character.
1767  */
1768     int
1769 latin_ptr2char(char_u *p)
1770 {
1771     return *p;
1772 }
1773 
1774     static int
1775 dbcs_ptr2char(char_u *p)
1776 {
1777     if (MB_BYTE2LEN(*p) > 1 && p[1] != NUL)
1778 	return (p[0] << 8) + p[1];
1779     return *p;
1780 }
1781 
1782 /*
1783  * Convert a UTF-8 byte sequence to a wide character.
1784  * If the sequence is illegal or truncated by a NUL the first byte is
1785  * returned.
1786  * For an overlong sequence this may return zero.
1787  * Does not include composing characters, of course.
1788  */
1789     int
1790 utf_ptr2char(char_u *p)
1791 {
1792     int		len;
1793 
1794     if (p[0] < 0x80)	/* be quick for ASCII */
1795 	return p[0];
1796 
1797     len = utf8len_tab_zero[p[0]];
1798     if (len > 1 && (p[1] & 0xc0) == 0x80)
1799     {
1800 	if (len == 2)
1801 	    return ((p[0] & 0x1f) << 6) + (p[1] & 0x3f);
1802 	if ((p[2] & 0xc0) == 0x80)
1803 	{
1804 	    if (len == 3)
1805 		return ((p[0] & 0x0f) << 12) + ((p[1] & 0x3f) << 6)
1806 		    + (p[2] & 0x3f);
1807 	    if ((p[3] & 0xc0) == 0x80)
1808 	    {
1809 		if (len == 4)
1810 		    return ((p[0] & 0x07) << 18) + ((p[1] & 0x3f) << 12)
1811 			+ ((p[2] & 0x3f) << 6) + (p[3] & 0x3f);
1812 		if ((p[4] & 0xc0) == 0x80)
1813 		{
1814 		    if (len == 5)
1815 			return ((p[0] & 0x03) << 24) + ((p[1] & 0x3f) << 18)
1816 			    + ((p[2] & 0x3f) << 12) + ((p[3] & 0x3f) << 6)
1817 			    + (p[4] & 0x3f);
1818 		    if ((p[5] & 0xc0) == 0x80 && len == 6)
1819 			return ((p[0] & 0x01) << 30) + ((p[1] & 0x3f) << 24)
1820 			    + ((p[2] & 0x3f) << 18) + ((p[3] & 0x3f) << 12)
1821 			    + ((p[4] & 0x3f) << 6) + (p[5] & 0x3f);
1822 		}
1823 	    }
1824 	}
1825     }
1826     /* Illegal value, just return the first byte */
1827     return p[0];
1828 }
1829 
1830 /*
1831  * Convert a UTF-8 byte sequence to a wide character.
1832  * String is assumed to be terminated by NUL or after "n" bytes, whichever
1833  * comes first.
1834  * The function is safe in the sense that it never accesses memory beyond the
1835  * first "n" bytes of "s".
1836  *
1837  * On success, returns decoded codepoint, advances "s" to the beginning of
1838  * next character and decreases "n" accordingly.
1839  *
1840  * If end of string was reached, returns 0 and, if "n" > 0, advances "s" past
1841  * NUL byte.
1842  *
1843  * If byte sequence is illegal or incomplete, returns -1 and does not advance
1844  * "s".
1845  */
1846     static int
1847 utf_safe_read_char_adv(char_u **s, size_t *n)
1848 {
1849     int		c, k;
1850 
1851     if (*n == 0) /* end of buffer */
1852 	return 0;
1853 
1854     k = utf8len_tab_zero[**s];
1855 
1856     if (k == 1)
1857     {
1858 	/* ASCII character or NUL */
1859 	(*n)--;
1860 	return *(*s)++;
1861     }
1862 
1863     if ((size_t)k <= *n)
1864     {
1865 	/* We have a multibyte sequence and it isn't truncated by buffer
1866 	 * limits so utf_ptr2char() is safe to use. Or the first byte is
1867 	 * illegal (k=0), and it's also safe to use utf_ptr2char(). */
1868 	c = utf_ptr2char(*s);
1869 
1870 	/* On failure, utf_ptr2char() returns the first byte, so here we
1871 	 * check equality with the first byte. The only non-ASCII character
1872 	 * which equals the first byte of its own UTF-8 representation is
1873 	 * U+00C3 (UTF-8: 0xC3 0x83), so need to check that special case too.
1874 	 * It's safe even if n=1, else we would have k=2 > n. */
1875 	if (c != (int)(**s) || (c == 0xC3 && (*s)[1] == 0x83))
1876 	{
1877 	    /* byte sequence was successfully decoded */
1878 	    *s += k;
1879 	    *n -= k;
1880 	    return c;
1881 	}
1882     }
1883 
1884     /* byte sequence is incomplete or illegal */
1885     return -1;
1886 }
1887 
1888 /*
1889  * Get character at **pp and advance *pp to the next character.
1890  * Note: composing characters are skipped!
1891  */
1892     int
1893 mb_ptr2char_adv(char_u **pp)
1894 {
1895     int		c;
1896 
1897     c = (*mb_ptr2char)(*pp);
1898     *pp += (*mb_ptr2len)(*pp);
1899     return c;
1900 }
1901 
1902 /*
1903  * Get character at **pp and advance *pp to the next character.
1904  * Note: composing characters are returned as separate characters.
1905  */
1906     int
1907 mb_cptr2char_adv(char_u **pp)
1908 {
1909     int		c;
1910 
1911     c = (*mb_ptr2char)(*pp);
1912     if (enc_utf8)
1913 	*pp += utf_ptr2len(*pp);
1914     else
1915 	*pp += (*mb_ptr2len)(*pp);
1916     return c;
1917 }
1918 
1919 #if defined(FEAT_ARABIC) || defined(PROTO)
1920 /*
1921  * Check if the character pointed to by "p2" is a composing character when it
1922  * comes after "p1".  For Arabic sometimes "ab" is replaced with "c", which
1923  * behaves like a composing character.
1924  */
1925     int
1926 utf_composinglike(char_u *p1, char_u *p2)
1927 {
1928     int		c2;
1929 
1930     c2 = utf_ptr2char(p2);
1931     if (utf_iscomposing(c2))
1932 	return TRUE;
1933     if (!arabic_maycombine(c2))
1934 	return FALSE;
1935     return arabic_combine(utf_ptr2char(p1), c2);
1936 }
1937 #endif
1938 
1939 /*
1940  * Convert a UTF-8 byte string to a wide character.  Also get up to MAX_MCO
1941  * composing characters.
1942  */
1943     int
1944 utfc_ptr2char(
1945     char_u	*p,
1946     int		*pcc)	/* return: composing chars, last one is 0 */
1947 {
1948     int		len;
1949     int		c;
1950     int		cc;
1951     int		i = 0;
1952 
1953     c = utf_ptr2char(p);
1954     len = utf_ptr2len(p);
1955 
1956     /* Only accept a composing char when the first char isn't illegal. */
1957     if ((len > 1 || *p < 0x80)
1958 	    && p[len] >= 0x80
1959 	    && UTF_COMPOSINGLIKE(p, p + len))
1960     {
1961 	cc = utf_ptr2char(p + len);
1962 	for (;;)
1963 	{
1964 	    pcc[i++] = cc;
1965 	    if (i == MAX_MCO)
1966 		break;
1967 	    len += utf_ptr2len(p + len);
1968 	    if (p[len] < 0x80 || !utf_iscomposing(cc = utf_ptr2char(p + len)))
1969 		break;
1970 	}
1971     }
1972 
1973     if (i < MAX_MCO)	/* last composing char must be 0 */
1974 	pcc[i] = 0;
1975 
1976     return c;
1977 }
1978 
1979 /*
1980  * Convert a UTF-8 byte string to a wide character.  Also get up to MAX_MCO
1981  * composing characters.  Use no more than p[maxlen].
1982  */
1983     int
1984 utfc_ptr2char_len(
1985     char_u	*p,
1986     int		*pcc,	/* return: composing chars, last one is 0 */
1987     int		maxlen)
1988 {
1989     int		len;
1990     int		c;
1991     int		cc;
1992     int		i = 0;
1993 
1994     c = utf_ptr2char(p);
1995     len = utf_ptr2len_len(p, maxlen);
1996     /* Only accept a composing char when the first char isn't illegal. */
1997     if ((len > 1 || *p < 0x80)
1998 	    && len < maxlen
1999 	    && p[len] >= 0x80
2000 	    && UTF_COMPOSINGLIKE(p, p + len))
2001     {
2002 	cc = utf_ptr2char(p + len);
2003 	for (;;)
2004 	{
2005 	    pcc[i++] = cc;
2006 	    if (i == MAX_MCO)
2007 		break;
2008 	    len += utf_ptr2len_len(p + len, maxlen - len);
2009 	    if (len >= maxlen
2010 		    || p[len] < 0x80
2011 		    || !utf_iscomposing(cc = utf_ptr2char(p + len)))
2012 		break;
2013 	}
2014     }
2015 
2016     if (i < MAX_MCO)	/* last composing char must be 0 */
2017 	pcc[i] = 0;
2018 
2019     return c;
2020 }
2021 
2022 /*
2023  * Convert the character at screen position "off" to a sequence of bytes.
2024  * Includes the composing characters.
2025  * "buf" must at least have the length MB_MAXBYTES + 1.
2026  * Only to be used when ScreenLinesUC[off] != 0.
2027  * Returns the produced number of bytes.
2028  */
2029     int
2030 utfc_char2bytes(int off, char_u *buf)
2031 {
2032     int		len;
2033     int		i;
2034 
2035     len = utf_char2bytes(ScreenLinesUC[off], buf);
2036     for (i = 0; i < Screen_mco; ++i)
2037     {
2038 	if (ScreenLinesC[i][off] == 0)
2039 	    break;
2040 	len += utf_char2bytes(ScreenLinesC[i][off], buf + len);
2041     }
2042     return len;
2043 }
2044 
2045 /*
2046  * Get the length of a UTF-8 byte sequence, not including any following
2047  * composing characters.
2048  * Returns 0 for "".
2049  * Returns 1 for an illegal byte sequence.
2050  */
2051     int
2052 utf_ptr2len(char_u *p)
2053 {
2054     int		len;
2055     int		i;
2056 
2057     if (*p == NUL)
2058 	return 0;
2059     len = utf8len_tab[*p];
2060     for (i = 1; i < len; ++i)
2061 	if ((p[i] & 0xc0) != 0x80)
2062 	    return 1;
2063     return len;
2064 }
2065 
2066 /*
2067  * Return length of UTF-8 character, obtained from the first byte.
2068  * "b" must be between 0 and 255!
2069  * Returns 1 for an invalid first byte value.
2070  */
2071     int
2072 utf_byte2len(int b)
2073 {
2074     return utf8len_tab[b];
2075 }
2076 
2077 /*
2078  * Get the length of UTF-8 byte sequence "p[size]".  Does not include any
2079  * following composing characters.
2080  * Returns 1 for "".
2081  * Returns 1 for an illegal byte sequence (also in incomplete byte seq.).
2082  * Returns number > "size" for an incomplete byte sequence.
2083  * Never returns zero.
2084  */
2085     int
2086 utf_ptr2len_len(char_u *p, int size)
2087 {
2088     int		len;
2089     int		i;
2090     int		m;
2091 
2092     len = utf8len_tab[*p];
2093     if (len == 1)
2094 	return 1;	/* NUL, ascii or illegal lead byte */
2095     if (len > size)
2096 	m = size;	/* incomplete byte sequence. */
2097     else
2098 	m = len;
2099     for (i = 1; i < m; ++i)
2100 	if ((p[i] & 0xc0) != 0x80)
2101 	    return 1;
2102     return len;
2103 }
2104 
2105 /*
2106  * Return the number of bytes the UTF-8 encoding of the character at "p" takes.
2107  * This includes following composing characters.
2108  */
2109     int
2110 utfc_ptr2len(char_u *p)
2111 {
2112     int		len;
2113     int		b0 = *p;
2114 #ifdef FEAT_ARABIC
2115     int		prevlen;
2116 #endif
2117 
2118     if (b0 == NUL)
2119 	return 0;
2120     if (b0 < 0x80 && p[1] < 0x80)	/* be quick for ASCII */
2121 	return 1;
2122 
2123     /* Skip over first UTF-8 char, stopping at a NUL byte. */
2124     len = utf_ptr2len(p);
2125 
2126     /* Check for illegal byte. */
2127     if (len == 1 && b0 >= 0x80)
2128 	return 1;
2129 
2130     /*
2131      * Check for composing characters.  We can handle only the first six, but
2132      * skip all of them (otherwise the cursor would get stuck).
2133      */
2134 #ifdef FEAT_ARABIC
2135     prevlen = 0;
2136 #endif
2137     for (;;)
2138     {
2139 	if (p[len] < 0x80 || !UTF_COMPOSINGLIKE(p + prevlen, p + len))
2140 	    return len;
2141 
2142 	/* Skip over composing char */
2143 #ifdef FEAT_ARABIC
2144 	prevlen = len;
2145 #endif
2146 	len += utf_ptr2len(p + len);
2147     }
2148 }
2149 
2150 /*
2151  * Return the number of bytes the UTF-8 encoding of the character at "p[size]"
2152  * takes.  This includes following composing characters.
2153  * Returns 0 for an empty string.
2154  * Returns 1 for an illegal char or an incomplete byte sequence.
2155  */
2156     int
2157 utfc_ptr2len_len(char_u *p, int size)
2158 {
2159     int		len;
2160 #ifdef FEAT_ARABIC
2161     int		prevlen;
2162 #endif
2163 
2164     if (size < 1 || *p == NUL)
2165 	return 0;
2166     if (p[0] < 0x80 && (size == 1 || p[1] < 0x80)) /* be quick for ASCII */
2167 	return 1;
2168 
2169     /* Skip over first UTF-8 char, stopping at a NUL byte. */
2170     len = utf_ptr2len_len(p, size);
2171 
2172     /* Check for illegal byte and incomplete byte sequence. */
2173     if ((len == 1 && p[0] >= 0x80) || len > size)
2174 	return 1;
2175 
2176     /*
2177      * Check for composing characters.  We can handle only the first six, but
2178      * skip all of them (otherwise the cursor would get stuck).
2179      */
2180 #ifdef FEAT_ARABIC
2181     prevlen = 0;
2182 #endif
2183     while (len < size)
2184     {
2185 	int	len_next_char;
2186 
2187 	if (p[len] < 0x80)
2188 	    break;
2189 
2190 	/*
2191 	 * Next character length should not go beyond size to ensure that
2192 	 * UTF_COMPOSINGLIKE(...) does not read beyond size.
2193 	 */
2194 	len_next_char = utf_ptr2len_len(p + len, size - len);
2195 	if (len_next_char > size - len)
2196 	    break;
2197 
2198 	if (!UTF_COMPOSINGLIKE(p + prevlen, p + len))
2199 	    break;
2200 
2201 	/* Skip over composing char */
2202 #ifdef FEAT_ARABIC
2203 	prevlen = len;
2204 #endif
2205 	len += len_next_char;
2206     }
2207     return len;
2208 }
2209 
2210 /*
2211  * Return the number of bytes the UTF-8 encoding of character "c" takes.
2212  * This does not include composing characters.
2213  */
2214     int
2215 utf_char2len(int c)
2216 {
2217     if (c < 0x80)
2218 	return 1;
2219     if (c < 0x800)
2220 	return 2;
2221     if (c < 0x10000)
2222 	return 3;
2223     if (c < 0x200000)
2224 	return 4;
2225     if (c < 0x4000000)
2226 	return 5;
2227     return 6;
2228 }
2229 
2230 /*
2231  * Convert Unicode character "c" to UTF-8 string in "buf[]".
2232  * Returns the number of bytes.
2233  */
2234     int
2235 utf_char2bytes(int c, char_u *buf)
2236 {
2237     if (c < 0x80)		/* 7 bits */
2238     {
2239 	buf[0] = c;
2240 	return 1;
2241     }
2242     if (c < 0x800)		/* 11 bits */
2243     {
2244 	buf[0] = 0xc0 + ((unsigned)c >> 6);
2245 	buf[1] = 0x80 + (c & 0x3f);
2246 	return 2;
2247     }
2248     if (c < 0x10000)		/* 16 bits */
2249     {
2250 	buf[0] = 0xe0 + ((unsigned)c >> 12);
2251 	buf[1] = 0x80 + (((unsigned)c >> 6) & 0x3f);
2252 	buf[2] = 0x80 + (c & 0x3f);
2253 	return 3;
2254     }
2255     if (c < 0x200000)		/* 21 bits */
2256     {
2257 	buf[0] = 0xf0 + ((unsigned)c >> 18);
2258 	buf[1] = 0x80 + (((unsigned)c >> 12) & 0x3f);
2259 	buf[2] = 0x80 + (((unsigned)c >> 6) & 0x3f);
2260 	buf[3] = 0x80 + (c & 0x3f);
2261 	return 4;
2262     }
2263     if (c < 0x4000000)		/* 26 bits */
2264     {
2265 	buf[0] = 0xf8 + ((unsigned)c >> 24);
2266 	buf[1] = 0x80 + (((unsigned)c >> 18) & 0x3f);
2267 	buf[2] = 0x80 + (((unsigned)c >> 12) & 0x3f);
2268 	buf[3] = 0x80 + (((unsigned)c >> 6) & 0x3f);
2269 	buf[4] = 0x80 + (c & 0x3f);
2270 	return 5;
2271     }
2272 				/* 31 bits */
2273     buf[0] = 0xfc + ((unsigned)c >> 30);
2274     buf[1] = 0x80 + (((unsigned)c >> 24) & 0x3f);
2275     buf[2] = 0x80 + (((unsigned)c >> 18) & 0x3f);
2276     buf[3] = 0x80 + (((unsigned)c >> 12) & 0x3f);
2277     buf[4] = 0x80 + (((unsigned)c >> 6) & 0x3f);
2278     buf[5] = 0x80 + (c & 0x3f);
2279     return 6;
2280 }
2281 
2282 #if defined(FEAT_TERMINAL) || defined(PROTO)
2283 /*
2284  * utf_iscomposing() with different argument type for libvterm.
2285  */
2286     int
2287 utf_iscomposing_uint(UINT32_T c)
2288 {
2289     return utf_iscomposing((int)c);
2290 }
2291 #endif
2292 
2293 /*
2294  * Return TRUE if "c" is a composing UTF-8 character.  This means it will be
2295  * drawn on top of the preceding character.
2296  * Based on code from Markus Kuhn.
2297  */
2298     int
2299 utf_iscomposing(int c)
2300 {
2301     /* Sorted list of non-overlapping intervals.
2302      * Generated by ../runtime/tools/unicode.vim. */
2303     static struct interval combining[] =
2304     {
2305 	{0x0300, 0x036f},
2306 	{0x0483, 0x0489},
2307 	{0x0591, 0x05bd},
2308 	{0x05bf, 0x05bf},
2309 	{0x05c1, 0x05c2},
2310 	{0x05c4, 0x05c5},
2311 	{0x05c7, 0x05c7},
2312 	{0x0610, 0x061a},
2313 	{0x064b, 0x065f},
2314 	{0x0670, 0x0670},
2315 	{0x06d6, 0x06dc},
2316 	{0x06df, 0x06e4},
2317 	{0x06e7, 0x06e8},
2318 	{0x06ea, 0x06ed},
2319 	{0x0711, 0x0711},
2320 	{0x0730, 0x074a},
2321 	{0x07a6, 0x07b0},
2322 	{0x07eb, 0x07f3},
2323 	{0x07fd, 0x07fd},
2324 	{0x0816, 0x0819},
2325 	{0x081b, 0x0823},
2326 	{0x0825, 0x0827},
2327 	{0x0829, 0x082d},
2328 	{0x0859, 0x085b},
2329 	{0x08d3, 0x08e1},
2330 	{0x08e3, 0x0903},
2331 	{0x093a, 0x093c},
2332 	{0x093e, 0x094f},
2333 	{0x0951, 0x0957},
2334 	{0x0962, 0x0963},
2335 	{0x0981, 0x0983},
2336 	{0x09bc, 0x09bc},
2337 	{0x09be, 0x09c4},
2338 	{0x09c7, 0x09c8},
2339 	{0x09cb, 0x09cd},
2340 	{0x09d7, 0x09d7},
2341 	{0x09e2, 0x09e3},
2342 	{0x09fe, 0x09fe},
2343 	{0x0a01, 0x0a03},
2344 	{0x0a3c, 0x0a3c},
2345 	{0x0a3e, 0x0a42},
2346 	{0x0a47, 0x0a48},
2347 	{0x0a4b, 0x0a4d},
2348 	{0x0a51, 0x0a51},
2349 	{0x0a70, 0x0a71},
2350 	{0x0a75, 0x0a75},
2351 	{0x0a81, 0x0a83},
2352 	{0x0abc, 0x0abc},
2353 	{0x0abe, 0x0ac5},
2354 	{0x0ac7, 0x0ac9},
2355 	{0x0acb, 0x0acd},
2356 	{0x0ae2, 0x0ae3},
2357 	{0x0afa, 0x0aff},
2358 	{0x0b01, 0x0b03},
2359 	{0x0b3c, 0x0b3c},
2360 	{0x0b3e, 0x0b44},
2361 	{0x0b47, 0x0b48},
2362 	{0x0b4b, 0x0b4d},
2363 	{0x0b56, 0x0b57},
2364 	{0x0b62, 0x0b63},
2365 	{0x0b82, 0x0b82},
2366 	{0x0bbe, 0x0bc2},
2367 	{0x0bc6, 0x0bc8},
2368 	{0x0bca, 0x0bcd},
2369 	{0x0bd7, 0x0bd7},
2370 	{0x0c00, 0x0c04},
2371 	{0x0c3e, 0x0c44},
2372 	{0x0c46, 0x0c48},
2373 	{0x0c4a, 0x0c4d},
2374 	{0x0c55, 0x0c56},
2375 	{0x0c62, 0x0c63},
2376 	{0x0c81, 0x0c83},
2377 	{0x0cbc, 0x0cbc},
2378 	{0x0cbe, 0x0cc4},
2379 	{0x0cc6, 0x0cc8},
2380 	{0x0cca, 0x0ccd},
2381 	{0x0cd5, 0x0cd6},
2382 	{0x0ce2, 0x0ce3},
2383 	{0x0d00, 0x0d03},
2384 	{0x0d3b, 0x0d3c},
2385 	{0x0d3e, 0x0d44},
2386 	{0x0d46, 0x0d48},
2387 	{0x0d4a, 0x0d4d},
2388 	{0x0d57, 0x0d57},
2389 	{0x0d62, 0x0d63},
2390 	{0x0d82, 0x0d83},
2391 	{0x0dca, 0x0dca},
2392 	{0x0dcf, 0x0dd4},
2393 	{0x0dd6, 0x0dd6},
2394 	{0x0dd8, 0x0ddf},
2395 	{0x0df2, 0x0df3},
2396 	{0x0e31, 0x0e31},
2397 	{0x0e34, 0x0e3a},
2398 	{0x0e47, 0x0e4e},
2399 	{0x0eb1, 0x0eb1},
2400 	{0x0eb4, 0x0ebc},
2401 	{0x0ec8, 0x0ecd},
2402 	{0x0f18, 0x0f19},
2403 	{0x0f35, 0x0f35},
2404 	{0x0f37, 0x0f37},
2405 	{0x0f39, 0x0f39},
2406 	{0x0f3e, 0x0f3f},
2407 	{0x0f71, 0x0f84},
2408 	{0x0f86, 0x0f87},
2409 	{0x0f8d, 0x0f97},
2410 	{0x0f99, 0x0fbc},
2411 	{0x0fc6, 0x0fc6},
2412 	{0x102b, 0x103e},
2413 	{0x1056, 0x1059},
2414 	{0x105e, 0x1060},
2415 	{0x1062, 0x1064},
2416 	{0x1067, 0x106d},
2417 	{0x1071, 0x1074},
2418 	{0x1082, 0x108d},
2419 	{0x108f, 0x108f},
2420 	{0x109a, 0x109d},
2421 	{0x135d, 0x135f},
2422 	{0x1712, 0x1714},
2423 	{0x1732, 0x1734},
2424 	{0x1752, 0x1753},
2425 	{0x1772, 0x1773},
2426 	{0x17b4, 0x17d3},
2427 	{0x17dd, 0x17dd},
2428 	{0x180b, 0x180d},
2429 	{0x1885, 0x1886},
2430 	{0x18a9, 0x18a9},
2431 	{0x1920, 0x192b},
2432 	{0x1930, 0x193b},
2433 	{0x1a17, 0x1a1b},
2434 	{0x1a55, 0x1a5e},
2435 	{0x1a60, 0x1a7c},
2436 	{0x1a7f, 0x1a7f},
2437 	{0x1ab0, 0x1abe},
2438 	{0x1b00, 0x1b04},
2439 	{0x1b34, 0x1b44},
2440 	{0x1b6b, 0x1b73},
2441 	{0x1b80, 0x1b82},
2442 	{0x1ba1, 0x1bad},
2443 	{0x1be6, 0x1bf3},
2444 	{0x1c24, 0x1c37},
2445 	{0x1cd0, 0x1cd2},
2446 	{0x1cd4, 0x1ce8},
2447 	{0x1ced, 0x1ced},
2448 	{0x1cf4, 0x1cf4},
2449 	{0x1cf7, 0x1cf9},
2450 	{0x1dc0, 0x1df9},
2451 	{0x1dfb, 0x1dff},
2452 	{0x20d0, 0x20f0},
2453 	{0x2cef, 0x2cf1},
2454 	{0x2d7f, 0x2d7f},
2455 	{0x2de0, 0x2dff},
2456 	{0x302a, 0x302f},
2457 	{0x3099, 0x309a},
2458 	{0xa66f, 0xa672},
2459 	{0xa674, 0xa67d},
2460 	{0xa69e, 0xa69f},
2461 	{0xa6f0, 0xa6f1},
2462 	{0xa802, 0xa802},
2463 	{0xa806, 0xa806},
2464 	{0xa80b, 0xa80b},
2465 	{0xa823, 0xa827},
2466 	{0xa880, 0xa881},
2467 	{0xa8b4, 0xa8c5},
2468 	{0xa8e0, 0xa8f1},
2469 	{0xa8ff, 0xa8ff},
2470 	{0xa926, 0xa92d},
2471 	{0xa947, 0xa953},
2472 	{0xa980, 0xa983},
2473 	{0xa9b3, 0xa9c0},
2474 	{0xa9e5, 0xa9e5},
2475 	{0xaa29, 0xaa36},
2476 	{0xaa43, 0xaa43},
2477 	{0xaa4c, 0xaa4d},
2478 	{0xaa7b, 0xaa7d},
2479 	{0xaab0, 0xaab0},
2480 	{0xaab2, 0xaab4},
2481 	{0xaab7, 0xaab8},
2482 	{0xaabe, 0xaabf},
2483 	{0xaac1, 0xaac1},
2484 	{0xaaeb, 0xaaef},
2485 	{0xaaf5, 0xaaf6},
2486 	{0xabe3, 0xabea},
2487 	{0xabec, 0xabed},
2488 	{0xfb1e, 0xfb1e},
2489 	{0xfe00, 0xfe0f},
2490 	{0xfe20, 0xfe2f},
2491 	{0x101fd, 0x101fd},
2492 	{0x102e0, 0x102e0},
2493 	{0x10376, 0x1037a},
2494 	{0x10a01, 0x10a03},
2495 	{0x10a05, 0x10a06},
2496 	{0x10a0c, 0x10a0f},
2497 	{0x10a38, 0x10a3a},
2498 	{0x10a3f, 0x10a3f},
2499 	{0x10ae5, 0x10ae6},
2500 	{0x10d24, 0x10d27},
2501 	{0x10f46, 0x10f50},
2502 	{0x11000, 0x11002},
2503 	{0x11038, 0x11046},
2504 	{0x1107f, 0x11082},
2505 	{0x110b0, 0x110ba},
2506 	{0x11100, 0x11102},
2507 	{0x11127, 0x11134},
2508 	{0x11145, 0x11146},
2509 	{0x11173, 0x11173},
2510 	{0x11180, 0x11182},
2511 	{0x111b3, 0x111c0},
2512 	{0x111c9, 0x111cc},
2513 	{0x1122c, 0x11237},
2514 	{0x1123e, 0x1123e},
2515 	{0x112df, 0x112ea},
2516 	{0x11300, 0x11303},
2517 	{0x1133b, 0x1133c},
2518 	{0x1133e, 0x11344},
2519 	{0x11347, 0x11348},
2520 	{0x1134b, 0x1134d},
2521 	{0x11357, 0x11357},
2522 	{0x11362, 0x11363},
2523 	{0x11366, 0x1136c},
2524 	{0x11370, 0x11374},
2525 	{0x11435, 0x11446},
2526 	{0x1145e, 0x1145e},
2527 	{0x114b0, 0x114c3},
2528 	{0x115af, 0x115b5},
2529 	{0x115b8, 0x115c0},
2530 	{0x115dc, 0x115dd},
2531 	{0x11630, 0x11640},
2532 	{0x116ab, 0x116b7},
2533 	{0x1171d, 0x1172b},
2534 	{0x1182c, 0x1183a},
2535 	{0x119d1, 0x119d7},
2536 	{0x119da, 0x119e0},
2537 	{0x119e4, 0x119e4},
2538 	{0x11a01, 0x11a0a},
2539 	{0x11a33, 0x11a39},
2540 	{0x11a3b, 0x11a3e},
2541 	{0x11a47, 0x11a47},
2542 	{0x11a51, 0x11a5b},
2543 	{0x11a8a, 0x11a99},
2544 	{0x11c2f, 0x11c36},
2545 	{0x11c38, 0x11c3f},
2546 	{0x11c92, 0x11ca7},
2547 	{0x11ca9, 0x11cb6},
2548 	{0x11d31, 0x11d36},
2549 	{0x11d3a, 0x11d3a},
2550 	{0x11d3c, 0x11d3d},
2551 	{0x11d3f, 0x11d45},
2552 	{0x11d47, 0x11d47},
2553 	{0x11d8a, 0x11d8e},
2554 	{0x11d90, 0x11d91},
2555 	{0x11d93, 0x11d97},
2556 	{0x11ef3, 0x11ef6},
2557 	{0x16af0, 0x16af4},
2558 	{0x16b30, 0x16b36},
2559 	{0x16f4f, 0x16f4f},
2560 	{0x16f51, 0x16f87},
2561 	{0x16f8f, 0x16f92},
2562 	{0x1bc9d, 0x1bc9e},
2563 	{0x1d165, 0x1d169},
2564 	{0x1d16d, 0x1d172},
2565 	{0x1d17b, 0x1d182},
2566 	{0x1d185, 0x1d18b},
2567 	{0x1d1aa, 0x1d1ad},
2568 	{0x1d242, 0x1d244},
2569 	{0x1da00, 0x1da36},
2570 	{0x1da3b, 0x1da6c},
2571 	{0x1da75, 0x1da75},
2572 	{0x1da84, 0x1da84},
2573 	{0x1da9b, 0x1da9f},
2574 	{0x1daa1, 0x1daaf},
2575 	{0x1e000, 0x1e006},
2576 	{0x1e008, 0x1e018},
2577 	{0x1e01b, 0x1e021},
2578 	{0x1e023, 0x1e024},
2579 	{0x1e026, 0x1e02a},
2580 	{0x1e130, 0x1e136},
2581 	{0x1e2ec, 0x1e2ef},
2582 	{0x1e8d0, 0x1e8d6},
2583 	{0x1e944, 0x1e94a},
2584 	{0xe0100, 0xe01ef}
2585     };
2586 
2587     return intable(combining, sizeof(combining), c);
2588 }
2589 
2590 /*
2591  * Return TRUE for characters that can be displayed in a normal way.
2592  * Only for characters of 0x100 and above!
2593  */
2594     int
2595 utf_printable(int c)
2596 {
2597 #ifdef USE_WCHAR_FUNCTIONS
2598     /*
2599      * Assume the iswprint() library function works better than our own stuff.
2600      */
2601     return iswprint(c);
2602 #else
2603     /* Sorted list of non-overlapping intervals.
2604      * 0xd800-0xdfff is reserved for UTF-16, actually illegal. */
2605     static struct interval nonprint[] =
2606     {
2607 	{0x070f, 0x070f}, {0x180b, 0x180e}, {0x200b, 0x200f}, {0x202a, 0x202e},
2608 	{0x206a, 0x206f}, {0xd800, 0xdfff}, {0xfeff, 0xfeff}, {0xfff9, 0xfffb},
2609 	{0xfffe, 0xffff}
2610     };
2611 
2612     return !intable(nonprint, sizeof(nonprint), c);
2613 #endif
2614 }
2615 
2616 /* Sorted list of non-overlapping intervals of all Emoji characters,
2617  * based on http://unicode.org/emoji/charts/emoji-list.html */
2618 static struct interval emoji_all[] =
2619 {
2620     {0x203c, 0x203c},
2621     {0x2049, 0x2049},
2622     {0x2122, 0x2122},
2623     {0x2139, 0x2139},
2624     {0x2194, 0x2199},
2625     {0x21a9, 0x21aa},
2626     {0x231a, 0x231b},
2627     {0x2328, 0x2328},
2628     {0x23cf, 0x23cf},
2629     {0x23e9, 0x23f3},
2630     {0x23f8, 0x23fa},
2631     {0x24c2, 0x24c2},
2632     {0x25aa, 0x25ab},
2633     {0x25b6, 0x25b6},
2634     {0x25c0, 0x25c0},
2635     {0x25fb, 0x25fe},
2636     {0x2600, 0x2604},
2637     {0x260e, 0x260e},
2638     {0x2611, 0x2611},
2639     {0x2614, 0x2615},
2640     {0x2618, 0x2618},
2641     {0x261d, 0x261d},
2642     {0x2620, 0x2620},
2643     {0x2622, 0x2623},
2644     {0x2626, 0x2626},
2645     {0x262a, 0x262a},
2646     {0x262e, 0x262f},
2647     {0x2638, 0x263a},
2648     {0x2640, 0x2640},
2649     {0x2642, 0x2642},
2650     {0x2648, 0x2653},
2651     {0x265f, 0x2660},
2652     {0x2663, 0x2663},
2653     {0x2665, 0x2666},
2654     {0x2668, 0x2668},
2655     {0x267b, 0x267b},
2656     {0x267e, 0x267f},
2657     {0x2692, 0x2697},
2658     {0x2699, 0x2699},
2659     {0x269b, 0x269c},
2660     {0x26a0, 0x26a1},
2661     {0x26aa, 0x26ab},
2662     {0x26b0, 0x26b1},
2663     {0x26bd, 0x26be},
2664     {0x26c4, 0x26c5},
2665     {0x26c8, 0x26c8},
2666     {0x26ce, 0x26cf},
2667     {0x26d1, 0x26d1},
2668     {0x26d3, 0x26d4},
2669     {0x26e9, 0x26ea},
2670     {0x26f0, 0x26f5},
2671     {0x26f7, 0x26fa},
2672     {0x26fd, 0x26fd},
2673     {0x2702, 0x2702},
2674     {0x2705, 0x2705},
2675     {0x2708, 0x270d},
2676     {0x270f, 0x270f},
2677     {0x2712, 0x2712},
2678     {0x2714, 0x2714},
2679     {0x2716, 0x2716},
2680     {0x271d, 0x271d},
2681     {0x2721, 0x2721},
2682     {0x2728, 0x2728},
2683     {0x2733, 0x2734},
2684     {0x2744, 0x2744},
2685     {0x2747, 0x2747},
2686     {0x274c, 0x274c},
2687     {0x274e, 0x274e},
2688     {0x2753, 0x2755},
2689     {0x2757, 0x2757},
2690     {0x2763, 0x2764},
2691     {0x2795, 0x2797},
2692     {0x27a1, 0x27a1},
2693     {0x27b0, 0x27b0},
2694     {0x27bf, 0x27bf},
2695     {0x2934, 0x2935},
2696     {0x2b05, 0x2b07},
2697     {0x2b1b, 0x2b1c},
2698     {0x2b50, 0x2b50},
2699     {0x2b55, 0x2b55},
2700     {0x3030, 0x3030},
2701     {0x303d, 0x303d},
2702     {0x3297, 0x3297},
2703     {0x3299, 0x3299},
2704     {0x1f004, 0x1f004},
2705     {0x1f0cf, 0x1f0cf},
2706     {0x1f170, 0x1f171},
2707     {0x1f17e, 0x1f17f},
2708     {0x1f18e, 0x1f18e},
2709     {0x1f191, 0x1f19a},
2710     {0x1f1e6, 0x1f1ff},
2711     {0x1f201, 0x1f202},
2712     {0x1f21a, 0x1f21a},
2713     {0x1f22f, 0x1f22f},
2714     {0x1f232, 0x1f23a},
2715     {0x1f250, 0x1f251},
2716     {0x1f300, 0x1f321},
2717     {0x1f324, 0x1f393},
2718     {0x1f396, 0x1f397},
2719     {0x1f399, 0x1f39b},
2720     {0x1f39e, 0x1f3f0},
2721     {0x1f3f3, 0x1f3f5},
2722     {0x1f3f7, 0x1f4fd},
2723     {0x1f4ff, 0x1f53d},
2724     {0x1f549, 0x1f54e},
2725     {0x1f550, 0x1f567},
2726     {0x1f56f, 0x1f570},
2727     {0x1f573, 0x1f57a},
2728     {0x1f587, 0x1f587},
2729     {0x1f58a, 0x1f58d},
2730     {0x1f590, 0x1f590},
2731     {0x1f595, 0x1f596},
2732     {0x1f5a4, 0x1f5a5},
2733     {0x1f5a8, 0x1f5a8},
2734     {0x1f5b1, 0x1f5b2},
2735     {0x1f5bc, 0x1f5bc},
2736     {0x1f5c2, 0x1f5c4},
2737     {0x1f5d1, 0x1f5d3},
2738     {0x1f5dc, 0x1f5de},
2739     {0x1f5e1, 0x1f5e1},
2740     {0x1f5e3, 0x1f5e3},
2741     {0x1f5e8, 0x1f5e8},
2742     {0x1f5ef, 0x1f5ef},
2743     {0x1f5f3, 0x1f5f3},
2744     {0x1f5fa, 0x1f64f},
2745     {0x1f680, 0x1f6c5},
2746     {0x1f6cb, 0x1f6d2},
2747     {0x1f6e0, 0x1f6e5},
2748     {0x1f6e9, 0x1f6e9},
2749     {0x1f6eb, 0x1f6ec},
2750     {0x1f6f0, 0x1f6f0},
2751     {0x1f6f3, 0x1f6f9},
2752     {0x1f910, 0x1f93a},
2753     {0x1f93c, 0x1f93e},
2754     {0x1f940, 0x1f945},
2755     {0x1f947, 0x1f970},
2756     {0x1f973, 0x1f976},
2757     {0x1f97a, 0x1f97a},
2758     {0x1f97c, 0x1f9a2},
2759     {0x1f9b0, 0x1f9b9},
2760     {0x1f9c0, 0x1f9c2},
2761     {0x1f9d0, 0x1f9ff}
2762 };
2763 
2764 /*
2765  * Get class of a Unicode character.
2766  * 0: white space
2767  * 1: punctuation
2768  * 2 or bigger: some class of word character.
2769  */
2770     int
2771 utf_class(int c)
2772 {
2773     return utf_class_buf(c, curbuf);
2774 }
2775 
2776     int
2777 utf_class_buf(int c, buf_T *buf)
2778 {
2779     /* sorted list of non-overlapping intervals */
2780     static struct clinterval
2781     {
2782 	unsigned int first;
2783 	unsigned int last;
2784 	unsigned int class;
2785     } classes[] =
2786     {
2787 	{0x037e, 0x037e, 1},		/* Greek question mark */
2788 	{0x0387, 0x0387, 1},		/* Greek ano teleia */
2789 	{0x055a, 0x055f, 1},		/* Armenian punctuation */
2790 	{0x0589, 0x0589, 1},		/* Armenian full stop */
2791 	{0x05be, 0x05be, 1},
2792 	{0x05c0, 0x05c0, 1},
2793 	{0x05c3, 0x05c3, 1},
2794 	{0x05f3, 0x05f4, 1},
2795 	{0x060c, 0x060c, 1},
2796 	{0x061b, 0x061b, 1},
2797 	{0x061f, 0x061f, 1},
2798 	{0x066a, 0x066d, 1},
2799 	{0x06d4, 0x06d4, 1},
2800 	{0x0700, 0x070d, 1},		/* Syriac punctuation */
2801 	{0x0964, 0x0965, 1},
2802 	{0x0970, 0x0970, 1},
2803 	{0x0df4, 0x0df4, 1},
2804 	{0x0e4f, 0x0e4f, 1},
2805 	{0x0e5a, 0x0e5b, 1},
2806 	{0x0f04, 0x0f12, 1},
2807 	{0x0f3a, 0x0f3d, 1},
2808 	{0x0f85, 0x0f85, 1},
2809 	{0x104a, 0x104f, 1},		/* Myanmar punctuation */
2810 	{0x10fb, 0x10fb, 1},		/* Georgian punctuation */
2811 	{0x1361, 0x1368, 1},		/* Ethiopic punctuation */
2812 	{0x166d, 0x166e, 1},		/* Canadian Syl. punctuation */
2813 	{0x1680, 0x1680, 0},
2814 	{0x169b, 0x169c, 1},
2815 	{0x16eb, 0x16ed, 1},
2816 	{0x1735, 0x1736, 1},
2817 	{0x17d4, 0x17dc, 1},		/* Khmer punctuation */
2818 	{0x1800, 0x180a, 1},		/* Mongolian punctuation */
2819 	{0x2000, 0x200b, 0},		/* spaces */
2820 	{0x200c, 0x2027, 1},		/* punctuation and symbols */
2821 	{0x2028, 0x2029, 0},
2822 	{0x202a, 0x202e, 1},		/* punctuation and symbols */
2823 	{0x202f, 0x202f, 0},
2824 	{0x2030, 0x205e, 1},		/* punctuation and symbols */
2825 	{0x205f, 0x205f, 0},
2826 	{0x2060, 0x27ff, 1},		/* punctuation and symbols */
2827 	{0x2070, 0x207f, 0x2070},	/* superscript */
2828 	{0x2080, 0x2094, 0x2080},	/* subscript */
2829 	{0x20a0, 0x27ff, 1},		/* all kinds of symbols */
2830 	{0x2800, 0x28ff, 0x2800},	/* braille */
2831 	{0x2900, 0x2998, 1},		/* arrows, brackets, etc. */
2832 	{0x29d8, 0x29db, 1},
2833 	{0x29fc, 0x29fd, 1},
2834 	{0x2e00, 0x2e7f, 1},		/* supplemental punctuation */
2835 	{0x3000, 0x3000, 0},		/* ideographic space */
2836 	{0x3001, 0x3020, 1},		/* ideographic punctuation */
2837 	{0x3030, 0x3030, 1},
2838 	{0x303d, 0x303d, 1},
2839 	{0x3040, 0x309f, 0x3040},	/* Hiragana */
2840 	{0x30a0, 0x30ff, 0x30a0},	/* Katakana */
2841 	{0x3300, 0x9fff, 0x4e00},	/* CJK Ideographs */
2842 	{0xac00, 0xd7a3, 0xac00},	/* Hangul Syllables */
2843 	{0xf900, 0xfaff, 0x4e00},	/* CJK Ideographs */
2844 	{0xfd3e, 0xfd3f, 1},
2845 	{0xfe30, 0xfe6b, 1},		/* punctuation forms */
2846 	{0xff00, 0xff0f, 1},		/* half/fullwidth ASCII */
2847 	{0xff1a, 0xff20, 1},		/* half/fullwidth ASCII */
2848 	{0xff3b, 0xff40, 1},		/* half/fullwidth ASCII */
2849 	{0xff5b, 0xff65, 1},		/* half/fullwidth ASCII */
2850 	{0x1d000, 0x1d24f, 1},		/* Musical notation */
2851 	{0x1d400, 0x1d7ff, 1},		/* Mathematical Alphanumeric Symbols */
2852 	{0x1f000, 0x1f2ff, 1},		/* Game pieces; enclosed characters */
2853 	{0x1f300, 0x1f9ff, 1},		/* Many symbol blocks */
2854 	{0x20000, 0x2a6df, 0x4e00},	/* CJK Ideographs */
2855 	{0x2a700, 0x2b73f, 0x4e00},	/* CJK Ideographs */
2856 	{0x2b740, 0x2b81f, 0x4e00},	/* CJK Ideographs */
2857 	{0x2f800, 0x2fa1f, 0x4e00},	/* CJK Ideographs */
2858     };
2859 
2860     int bot = 0;
2861     int top = sizeof(classes) / sizeof(struct clinterval) - 1;
2862     int mid;
2863 
2864     /* First quick check for Latin1 characters, use 'iskeyword'. */
2865     if (c < 0x100)
2866     {
2867 	if (c == ' ' || c == '\t' || c == NUL || c == 0xa0)
2868 	    return 0;	    /* blank */
2869 	if (vim_iswordc_buf(c, buf))
2870 	    return 2;	    /* word character */
2871 	return 1;	    /* punctuation */
2872     }
2873 
2874     /* binary search in table */
2875     while (top >= bot)
2876     {
2877 	mid = (bot + top) / 2;
2878 	if (classes[mid].last < (unsigned int)c)
2879 	    bot = mid + 1;
2880 	else if (classes[mid].first > (unsigned int)c)
2881 	    top = mid - 1;
2882 	else
2883 	    return (int)classes[mid].class;
2884     }
2885 
2886     /* emoji */
2887     if (intable(emoji_all, sizeof(emoji_all), c))
2888 	return 3;
2889 
2890     /* most other characters are "word" characters */
2891     return 2;
2892 }
2893 
2894     int
2895 utf_ambiguous_width(int c)
2896 {
2897     return c >= 0x80 && (intable(ambiguous, sizeof(ambiguous), c)
2898 	    || intable(emoji_all, sizeof(emoji_all), c));
2899 }
2900 
2901 /*
2902  * Code for Unicode case-dependent operations.  Based on notes in
2903  * http://www.unicode.org/Public/UNIDATA/CaseFolding.txt
2904  * This code uses simple case folding, not full case folding.
2905  * Last updated for Unicode 5.2.
2906  */
2907 
2908 /*
2909  * The following tables are built by ../runtime/tools/unicode.vim.
2910  * They must be in numeric order, because we use binary search.
2911  * An entry such as {0x41,0x5a,1,32} means that Unicode characters in the
2912  * range from 0x41 to 0x5a inclusive, stepping by 1, are changed to
2913  * folded/upper/lower by adding 32.
2914  */
2915 typedef struct
2916 {
2917     int rangeStart;
2918     int rangeEnd;
2919     int step;
2920     int offset;
2921 } convertStruct;
2922 
2923 static convertStruct foldCase[] =
2924 {
2925 	{0x41,0x5a,1,32},
2926 	{0xb5,0xb5,-1,775},
2927 	{0xc0,0xd6,1,32},
2928 	{0xd8,0xde,1,32},
2929 	{0x100,0x12e,2,1},
2930 	{0x132,0x136,2,1},
2931 	{0x139,0x147,2,1},
2932 	{0x14a,0x176,2,1},
2933 	{0x178,0x178,-1,-121},
2934 	{0x179,0x17d,2,1},
2935 	{0x17f,0x17f,-1,-268},
2936 	{0x181,0x181,-1,210},
2937 	{0x182,0x184,2,1},
2938 	{0x186,0x186,-1,206},
2939 	{0x187,0x187,-1,1},
2940 	{0x189,0x18a,1,205},
2941 	{0x18b,0x18b,-1,1},
2942 	{0x18e,0x18e,-1,79},
2943 	{0x18f,0x18f,-1,202},
2944 	{0x190,0x190,-1,203},
2945 	{0x191,0x191,-1,1},
2946 	{0x193,0x193,-1,205},
2947 	{0x194,0x194,-1,207},
2948 	{0x196,0x196,-1,211},
2949 	{0x197,0x197,-1,209},
2950 	{0x198,0x198,-1,1},
2951 	{0x19c,0x19c,-1,211},
2952 	{0x19d,0x19d,-1,213},
2953 	{0x19f,0x19f,-1,214},
2954 	{0x1a0,0x1a4,2,1},
2955 	{0x1a6,0x1a6,-1,218},
2956 	{0x1a7,0x1a7,-1,1},
2957 	{0x1a9,0x1a9,-1,218},
2958 	{0x1ac,0x1ac,-1,1},
2959 	{0x1ae,0x1ae,-1,218},
2960 	{0x1af,0x1af,-1,1},
2961 	{0x1b1,0x1b2,1,217},
2962 	{0x1b3,0x1b5,2,1},
2963 	{0x1b7,0x1b7,-1,219},
2964 	{0x1b8,0x1bc,4,1},
2965 	{0x1c4,0x1c4,-1,2},
2966 	{0x1c5,0x1c5,-1,1},
2967 	{0x1c7,0x1c7,-1,2},
2968 	{0x1c8,0x1c8,-1,1},
2969 	{0x1ca,0x1ca,-1,2},
2970 	{0x1cb,0x1db,2,1},
2971 	{0x1de,0x1ee,2,1},
2972 	{0x1f1,0x1f1,-1,2},
2973 	{0x1f2,0x1f4,2,1},
2974 	{0x1f6,0x1f6,-1,-97},
2975 	{0x1f7,0x1f7,-1,-56},
2976 	{0x1f8,0x21e,2,1},
2977 	{0x220,0x220,-1,-130},
2978 	{0x222,0x232,2,1},
2979 	{0x23a,0x23a,-1,10795},
2980 	{0x23b,0x23b,-1,1},
2981 	{0x23d,0x23d,-1,-163},
2982 	{0x23e,0x23e,-1,10792},
2983 	{0x241,0x241,-1,1},
2984 	{0x243,0x243,-1,-195},
2985 	{0x244,0x244,-1,69},
2986 	{0x245,0x245,-1,71},
2987 	{0x246,0x24e,2,1},
2988 	{0x345,0x345,-1,116},
2989 	{0x370,0x372,2,1},
2990 	{0x376,0x376,-1,1},
2991 	{0x37f,0x37f,-1,116},
2992 	{0x386,0x386,-1,38},
2993 	{0x388,0x38a,1,37},
2994 	{0x38c,0x38c,-1,64},
2995 	{0x38e,0x38f,1,63},
2996 	{0x391,0x3a1,1,32},
2997 	{0x3a3,0x3ab,1,32},
2998 	{0x3c2,0x3c2,-1,1},
2999 	{0x3cf,0x3cf,-1,8},
3000 	{0x3d0,0x3d0,-1,-30},
3001 	{0x3d1,0x3d1,-1,-25},
3002 	{0x3d5,0x3d5,-1,-15},
3003 	{0x3d6,0x3d6,-1,-22},
3004 	{0x3d8,0x3ee,2,1},
3005 	{0x3f0,0x3f0,-1,-54},
3006 	{0x3f1,0x3f1,-1,-48},
3007 	{0x3f4,0x3f4,-1,-60},
3008 	{0x3f5,0x3f5,-1,-64},
3009 	{0x3f7,0x3f7,-1,1},
3010 	{0x3f9,0x3f9,-1,-7},
3011 	{0x3fa,0x3fa,-1,1},
3012 	{0x3fd,0x3ff,1,-130},
3013 	{0x400,0x40f,1,80},
3014 	{0x410,0x42f,1,32},
3015 	{0x460,0x480,2,1},
3016 	{0x48a,0x4be,2,1},
3017 	{0x4c0,0x4c0,-1,15},
3018 	{0x4c1,0x4cd,2,1},
3019 	{0x4d0,0x52e,2,1},
3020 	{0x531,0x556,1,48},
3021 	{0x10a0,0x10c5,1,7264},
3022 	{0x10c7,0x10cd,6,7264},
3023 	{0x13f8,0x13fd,1,-8},
3024 	{0x1c80,0x1c80,-1,-6222},
3025 	{0x1c81,0x1c81,-1,-6221},
3026 	{0x1c82,0x1c82,-1,-6212},
3027 	{0x1c83,0x1c84,1,-6210},
3028 	{0x1c85,0x1c85,-1,-6211},
3029 	{0x1c86,0x1c86,-1,-6204},
3030 	{0x1c87,0x1c87,-1,-6180},
3031 	{0x1c88,0x1c88,-1,35267},
3032 	{0x1c90,0x1cba,1,-3008},
3033 	{0x1cbd,0x1cbf,1,-3008},
3034 	{0x1e00,0x1e94,2,1},
3035 	{0x1e9b,0x1e9b,-1,-58},
3036 	{0x1e9e,0x1e9e,-1,-7615},
3037 	{0x1ea0,0x1efe,2,1},
3038 	{0x1f08,0x1f0f,1,-8},
3039 	{0x1f18,0x1f1d,1,-8},
3040 	{0x1f28,0x1f2f,1,-8},
3041 	{0x1f38,0x1f3f,1,-8},
3042 	{0x1f48,0x1f4d,1,-8},
3043 	{0x1f59,0x1f5f,2,-8},
3044 	{0x1f68,0x1f6f,1,-8},
3045 	{0x1f88,0x1f8f,1,-8},
3046 	{0x1f98,0x1f9f,1,-8},
3047 	{0x1fa8,0x1faf,1,-8},
3048 	{0x1fb8,0x1fb9,1,-8},
3049 	{0x1fba,0x1fbb,1,-74},
3050 	{0x1fbc,0x1fbc,-1,-9},
3051 	{0x1fbe,0x1fbe,-1,-7173},
3052 	{0x1fc8,0x1fcb,1,-86},
3053 	{0x1fcc,0x1fcc,-1,-9},
3054 	{0x1fd8,0x1fd9,1,-8},
3055 	{0x1fda,0x1fdb,1,-100},
3056 	{0x1fe8,0x1fe9,1,-8},
3057 	{0x1fea,0x1feb,1,-112},
3058 	{0x1fec,0x1fec,-1,-7},
3059 	{0x1ff8,0x1ff9,1,-128},
3060 	{0x1ffa,0x1ffb,1,-126},
3061 	{0x1ffc,0x1ffc,-1,-9},
3062 	{0x2126,0x2126,-1,-7517},
3063 	{0x212a,0x212a,-1,-8383},
3064 	{0x212b,0x212b,-1,-8262},
3065 	{0x2132,0x2132,-1,28},
3066 	{0x2160,0x216f,1,16},
3067 	{0x2183,0x2183,-1,1},
3068 	{0x24b6,0x24cf,1,26},
3069 	{0x2c00,0x2c2e,1,48},
3070 	{0x2c60,0x2c60,-1,1},
3071 	{0x2c62,0x2c62,-1,-10743},
3072 	{0x2c63,0x2c63,-1,-3814},
3073 	{0x2c64,0x2c64,-1,-10727},
3074 	{0x2c67,0x2c6b,2,1},
3075 	{0x2c6d,0x2c6d,-1,-10780},
3076 	{0x2c6e,0x2c6e,-1,-10749},
3077 	{0x2c6f,0x2c6f,-1,-10783},
3078 	{0x2c70,0x2c70,-1,-10782},
3079 	{0x2c72,0x2c75,3,1},
3080 	{0x2c7e,0x2c7f,1,-10815},
3081 	{0x2c80,0x2ce2,2,1},
3082 	{0x2ceb,0x2ced,2,1},
3083 	{0x2cf2,0xa640,31054,1},
3084 	{0xa642,0xa66c,2,1},
3085 	{0xa680,0xa69a,2,1},
3086 	{0xa722,0xa72e,2,1},
3087 	{0xa732,0xa76e,2,1},
3088 	{0xa779,0xa77b,2,1},
3089 	{0xa77d,0xa77d,-1,-35332},
3090 	{0xa77e,0xa786,2,1},
3091 	{0xa78b,0xa78b,-1,1},
3092 	{0xa78d,0xa78d,-1,-42280},
3093 	{0xa790,0xa792,2,1},
3094 	{0xa796,0xa7a8,2,1},
3095 	{0xa7aa,0xa7aa,-1,-42308},
3096 	{0xa7ab,0xa7ab,-1,-42319},
3097 	{0xa7ac,0xa7ac,-1,-42315},
3098 	{0xa7ad,0xa7ad,-1,-42305},
3099 	{0xa7ae,0xa7ae,-1,-42308},
3100 	{0xa7b0,0xa7b0,-1,-42258},
3101 	{0xa7b1,0xa7b1,-1,-42282},
3102 	{0xa7b2,0xa7b2,-1,-42261},
3103 	{0xa7b3,0xa7b3,-1,928},
3104 	{0xa7b4,0xa7be,2,1},
3105 	{0xa7c2,0xa7c2,-1,1},
3106 	{0xa7c4,0xa7c4,-1,-48},
3107 	{0xa7c5,0xa7c5,-1,-42307},
3108 	{0xa7c6,0xa7c6,-1,-35384},
3109 	{0xab70,0xabbf,1,-38864},
3110 	{0xff21,0xff3a,1,32},
3111 	{0x10400,0x10427,1,40},
3112 	{0x104b0,0x104d3,1,40},
3113 	{0x10c80,0x10cb2,1,64},
3114 	{0x118a0,0x118bf,1,32},
3115 	{0x16e40,0x16e5f,1,32},
3116 	{0x1e900,0x1e921,1,34}
3117 };
3118 
3119 /*
3120  * Generic conversion function for case operations.
3121  * Return the converted equivalent of "a", which is a UCS-4 character.  Use
3122  * the given conversion "table".  Uses binary search on "table".
3123  */
3124     static int
3125 utf_convert(
3126     int			a,
3127     convertStruct	table[],
3128     int			tableSize)
3129 {
3130     int start, mid, end; /* indices into table */
3131     int entries = tableSize / sizeof(convertStruct);
3132 
3133     start = 0;
3134     end = entries;
3135     while (start < end)
3136     {
3137 	/* need to search further */
3138 	mid = (end + start) / 2;
3139 	if (table[mid].rangeEnd < a)
3140 	    start = mid + 1;
3141 	else
3142 	    end = mid;
3143     }
3144     if (start < entries
3145 	    && table[start].rangeStart <= a
3146 	    && a <= table[start].rangeEnd
3147 	    && (a - table[start].rangeStart) % table[start].step == 0)
3148 	return (a + table[start].offset);
3149     else
3150 	return a;
3151 }
3152 
3153 /*
3154  * Return the folded-case equivalent of "a", which is a UCS-4 character.  Uses
3155  * simple case folding.
3156  */
3157     int
3158 utf_fold(int a)
3159 {
3160     if (a < 0x80)
3161 	/* be fast for ASCII */
3162 	return a >= 0x41 && a <= 0x5a ? a + 32 : a;
3163     return utf_convert(a, foldCase, (int)sizeof(foldCase));
3164 }
3165 
3166 static convertStruct toLower[] =
3167 {
3168 	{0x41,0x5a,1,32},
3169 	{0xc0,0xd6,1,32},
3170 	{0xd8,0xde,1,32},
3171 	{0x100,0x12e,2,1},
3172 	{0x130,0x130,-1,-199},
3173 	{0x132,0x136,2,1},
3174 	{0x139,0x147,2,1},
3175 	{0x14a,0x176,2,1},
3176 	{0x178,0x178,-1,-121},
3177 	{0x179,0x17d,2,1},
3178 	{0x181,0x181,-1,210},
3179 	{0x182,0x184,2,1},
3180 	{0x186,0x186,-1,206},
3181 	{0x187,0x187,-1,1},
3182 	{0x189,0x18a,1,205},
3183 	{0x18b,0x18b,-1,1},
3184 	{0x18e,0x18e,-1,79},
3185 	{0x18f,0x18f,-1,202},
3186 	{0x190,0x190,-1,203},
3187 	{0x191,0x191,-1,1},
3188 	{0x193,0x193,-1,205},
3189 	{0x194,0x194,-1,207},
3190 	{0x196,0x196,-1,211},
3191 	{0x197,0x197,-1,209},
3192 	{0x198,0x198,-1,1},
3193 	{0x19c,0x19c,-1,211},
3194 	{0x19d,0x19d,-1,213},
3195 	{0x19f,0x19f,-1,214},
3196 	{0x1a0,0x1a4,2,1},
3197 	{0x1a6,0x1a6,-1,218},
3198 	{0x1a7,0x1a7,-1,1},
3199 	{0x1a9,0x1a9,-1,218},
3200 	{0x1ac,0x1ac,-1,1},
3201 	{0x1ae,0x1ae,-1,218},
3202 	{0x1af,0x1af,-1,1},
3203 	{0x1b1,0x1b2,1,217},
3204 	{0x1b3,0x1b5,2,1},
3205 	{0x1b7,0x1b7,-1,219},
3206 	{0x1b8,0x1bc,4,1},
3207 	{0x1c4,0x1c4,-1,2},
3208 	{0x1c5,0x1c5,-1,1},
3209 	{0x1c7,0x1c7,-1,2},
3210 	{0x1c8,0x1c8,-1,1},
3211 	{0x1ca,0x1ca,-1,2},
3212 	{0x1cb,0x1db,2,1},
3213 	{0x1de,0x1ee,2,1},
3214 	{0x1f1,0x1f1,-1,2},
3215 	{0x1f2,0x1f4,2,1},
3216 	{0x1f6,0x1f6,-1,-97},
3217 	{0x1f7,0x1f7,-1,-56},
3218 	{0x1f8,0x21e,2,1},
3219 	{0x220,0x220,-1,-130},
3220 	{0x222,0x232,2,1},
3221 	{0x23a,0x23a,-1,10795},
3222 	{0x23b,0x23b,-1,1},
3223 	{0x23d,0x23d,-1,-163},
3224 	{0x23e,0x23e,-1,10792},
3225 	{0x241,0x241,-1,1},
3226 	{0x243,0x243,-1,-195},
3227 	{0x244,0x244,-1,69},
3228 	{0x245,0x245,-1,71},
3229 	{0x246,0x24e,2,1},
3230 	{0x370,0x372,2,1},
3231 	{0x376,0x376,-1,1},
3232 	{0x37f,0x37f,-1,116},
3233 	{0x386,0x386,-1,38},
3234 	{0x388,0x38a,1,37},
3235 	{0x38c,0x38c,-1,64},
3236 	{0x38e,0x38f,1,63},
3237 	{0x391,0x3a1,1,32},
3238 	{0x3a3,0x3ab,1,32},
3239 	{0x3cf,0x3cf,-1,8},
3240 	{0x3d8,0x3ee,2,1},
3241 	{0x3f4,0x3f4,-1,-60},
3242 	{0x3f7,0x3f7,-1,1},
3243 	{0x3f9,0x3f9,-1,-7},
3244 	{0x3fa,0x3fa,-1,1},
3245 	{0x3fd,0x3ff,1,-130},
3246 	{0x400,0x40f,1,80},
3247 	{0x410,0x42f,1,32},
3248 	{0x460,0x480,2,1},
3249 	{0x48a,0x4be,2,1},
3250 	{0x4c0,0x4c0,-1,15},
3251 	{0x4c1,0x4cd,2,1},
3252 	{0x4d0,0x52e,2,1},
3253 	{0x531,0x556,1,48},
3254 	{0x10a0,0x10c5,1,7264},
3255 	{0x10c7,0x10cd,6,7264},
3256 	{0x13a0,0x13ef,1,38864},
3257 	{0x13f0,0x13f5,1,8},
3258 	{0x1c90,0x1cba,1,-3008},
3259 	{0x1cbd,0x1cbf,1,-3008},
3260 	{0x1e00,0x1e94,2,1},
3261 	{0x1e9e,0x1e9e,-1,-7615},
3262 	{0x1ea0,0x1efe,2,1},
3263 	{0x1f08,0x1f0f,1,-8},
3264 	{0x1f18,0x1f1d,1,-8},
3265 	{0x1f28,0x1f2f,1,-8},
3266 	{0x1f38,0x1f3f,1,-8},
3267 	{0x1f48,0x1f4d,1,-8},
3268 	{0x1f59,0x1f5f,2,-8},
3269 	{0x1f68,0x1f6f,1,-8},
3270 	{0x1f88,0x1f8f,1,-8},
3271 	{0x1f98,0x1f9f,1,-8},
3272 	{0x1fa8,0x1faf,1,-8},
3273 	{0x1fb8,0x1fb9,1,-8},
3274 	{0x1fba,0x1fbb,1,-74},
3275 	{0x1fbc,0x1fbc,-1,-9},
3276 	{0x1fc8,0x1fcb,1,-86},
3277 	{0x1fcc,0x1fcc,-1,-9},
3278 	{0x1fd8,0x1fd9,1,-8},
3279 	{0x1fda,0x1fdb,1,-100},
3280 	{0x1fe8,0x1fe9,1,-8},
3281 	{0x1fea,0x1feb,1,-112},
3282 	{0x1fec,0x1fec,-1,-7},
3283 	{0x1ff8,0x1ff9,1,-128},
3284 	{0x1ffa,0x1ffb,1,-126},
3285 	{0x1ffc,0x1ffc,-1,-9},
3286 	{0x2126,0x2126,-1,-7517},
3287 	{0x212a,0x212a,-1,-8383},
3288 	{0x212b,0x212b,-1,-8262},
3289 	{0x2132,0x2132,-1,28},
3290 	{0x2160,0x216f,1,16},
3291 	{0x2183,0x2183,-1,1},
3292 	{0x24b6,0x24cf,1,26},
3293 	{0x2c00,0x2c2e,1,48},
3294 	{0x2c60,0x2c60,-1,1},
3295 	{0x2c62,0x2c62,-1,-10743},
3296 	{0x2c63,0x2c63,-1,-3814},
3297 	{0x2c64,0x2c64,-1,-10727},
3298 	{0x2c67,0x2c6b,2,1},
3299 	{0x2c6d,0x2c6d,-1,-10780},
3300 	{0x2c6e,0x2c6e,-1,-10749},
3301 	{0x2c6f,0x2c6f,-1,-10783},
3302 	{0x2c70,0x2c70,-1,-10782},
3303 	{0x2c72,0x2c75,3,1},
3304 	{0x2c7e,0x2c7f,1,-10815},
3305 	{0x2c80,0x2ce2,2,1},
3306 	{0x2ceb,0x2ced,2,1},
3307 	{0x2cf2,0xa640,31054,1},
3308 	{0xa642,0xa66c,2,1},
3309 	{0xa680,0xa69a,2,1},
3310 	{0xa722,0xa72e,2,1},
3311 	{0xa732,0xa76e,2,1},
3312 	{0xa779,0xa77b,2,1},
3313 	{0xa77d,0xa77d,-1,-35332},
3314 	{0xa77e,0xa786,2,1},
3315 	{0xa78b,0xa78b,-1,1},
3316 	{0xa78d,0xa78d,-1,-42280},
3317 	{0xa790,0xa792,2,1},
3318 	{0xa796,0xa7a8,2,1},
3319 	{0xa7aa,0xa7aa,-1,-42308},
3320 	{0xa7ab,0xa7ab,-1,-42319},
3321 	{0xa7ac,0xa7ac,-1,-42315},
3322 	{0xa7ad,0xa7ad,-1,-42305},
3323 	{0xa7ae,0xa7ae,-1,-42308},
3324 	{0xa7b0,0xa7b0,-1,-42258},
3325 	{0xa7b1,0xa7b1,-1,-42282},
3326 	{0xa7b2,0xa7b2,-1,-42261},
3327 	{0xa7b3,0xa7b3,-1,928},
3328 	{0xa7b4,0xa7be,2,1},
3329 	{0xa7c2,0xa7c2,-1,1},
3330 	{0xa7c4,0xa7c4,-1,-48},
3331 	{0xa7c5,0xa7c5,-1,-42307},
3332 	{0xa7c6,0xa7c6,-1,-35384},
3333 	{0xff21,0xff3a,1,32},
3334 	{0x10400,0x10427,1,40},
3335 	{0x104b0,0x104d3,1,40},
3336 	{0x10c80,0x10cb2,1,64},
3337 	{0x118a0,0x118bf,1,32},
3338 	{0x16e40,0x16e5f,1,32},
3339 	{0x1e900,0x1e921,1,34}
3340 };
3341 
3342 static convertStruct toUpper[] =
3343 {
3344 	{0x61,0x7a,1,-32},
3345 	{0xb5,0xb5,-1,743},
3346 	{0xe0,0xf6,1,-32},
3347 	{0xf8,0xfe,1,-32},
3348 	{0xff,0xff,-1,121},
3349 	{0x101,0x12f,2,-1},
3350 	{0x131,0x131,-1,-232},
3351 	{0x133,0x137,2,-1},
3352 	{0x13a,0x148,2,-1},
3353 	{0x14b,0x177,2,-1},
3354 	{0x17a,0x17e,2,-1},
3355 	{0x17f,0x17f,-1,-300},
3356 	{0x180,0x180,-1,195},
3357 	{0x183,0x185,2,-1},
3358 	{0x188,0x18c,4,-1},
3359 	{0x192,0x192,-1,-1},
3360 	{0x195,0x195,-1,97},
3361 	{0x199,0x199,-1,-1},
3362 	{0x19a,0x19a,-1,163},
3363 	{0x19e,0x19e,-1,130},
3364 	{0x1a1,0x1a5,2,-1},
3365 	{0x1a8,0x1ad,5,-1},
3366 	{0x1b0,0x1b4,4,-1},
3367 	{0x1b6,0x1b9,3,-1},
3368 	{0x1bd,0x1bd,-1,-1},
3369 	{0x1bf,0x1bf,-1,56},
3370 	{0x1c5,0x1c5,-1,-1},
3371 	{0x1c6,0x1c6,-1,-2},
3372 	{0x1c8,0x1c8,-1,-1},
3373 	{0x1c9,0x1c9,-1,-2},
3374 	{0x1cb,0x1cb,-1,-1},
3375 	{0x1cc,0x1cc,-1,-2},
3376 	{0x1ce,0x1dc,2,-1},
3377 	{0x1dd,0x1dd,-1,-79},
3378 	{0x1df,0x1ef,2,-1},
3379 	{0x1f2,0x1f2,-1,-1},
3380 	{0x1f3,0x1f3,-1,-2},
3381 	{0x1f5,0x1f9,4,-1},
3382 	{0x1fb,0x21f,2,-1},
3383 	{0x223,0x233,2,-1},
3384 	{0x23c,0x23c,-1,-1},
3385 	{0x23f,0x240,1,10815},
3386 	{0x242,0x247,5,-1},
3387 	{0x249,0x24f,2,-1},
3388 	{0x250,0x250,-1,10783},
3389 	{0x251,0x251,-1,10780},
3390 	{0x252,0x252,-1,10782},
3391 	{0x253,0x253,-1,-210},
3392 	{0x254,0x254,-1,-206},
3393 	{0x256,0x257,1,-205},
3394 	{0x259,0x259,-1,-202},
3395 	{0x25b,0x25b,-1,-203},
3396 	{0x25c,0x25c,-1,42319},
3397 	{0x260,0x260,-1,-205},
3398 	{0x261,0x261,-1,42315},
3399 	{0x263,0x263,-1,-207},
3400 	{0x265,0x265,-1,42280},
3401 	{0x266,0x266,-1,42308},
3402 	{0x268,0x268,-1,-209},
3403 	{0x269,0x269,-1,-211},
3404 	{0x26a,0x26a,-1,42308},
3405 	{0x26b,0x26b,-1,10743},
3406 	{0x26c,0x26c,-1,42305},
3407 	{0x26f,0x26f,-1,-211},
3408 	{0x271,0x271,-1,10749},
3409 	{0x272,0x272,-1,-213},
3410 	{0x275,0x275,-1,-214},
3411 	{0x27d,0x27d,-1,10727},
3412 	{0x280,0x280,-1,-218},
3413 	{0x282,0x282,-1,42307},
3414 	{0x283,0x283,-1,-218},
3415 	{0x287,0x287,-1,42282},
3416 	{0x288,0x288,-1,-218},
3417 	{0x289,0x289,-1,-69},
3418 	{0x28a,0x28b,1,-217},
3419 	{0x28c,0x28c,-1,-71},
3420 	{0x292,0x292,-1,-219},
3421 	{0x29d,0x29d,-1,42261},
3422 	{0x29e,0x29e,-1,42258},
3423 	{0x345,0x345,-1,84},
3424 	{0x371,0x373,2,-1},
3425 	{0x377,0x377,-1,-1},
3426 	{0x37b,0x37d,1,130},
3427 	{0x3ac,0x3ac,-1,-38},
3428 	{0x3ad,0x3af,1,-37},
3429 	{0x3b1,0x3c1,1,-32},
3430 	{0x3c2,0x3c2,-1,-31},
3431 	{0x3c3,0x3cb,1,-32},
3432 	{0x3cc,0x3cc,-1,-64},
3433 	{0x3cd,0x3ce,1,-63},
3434 	{0x3d0,0x3d0,-1,-62},
3435 	{0x3d1,0x3d1,-1,-57},
3436 	{0x3d5,0x3d5,-1,-47},
3437 	{0x3d6,0x3d6,-1,-54},
3438 	{0x3d7,0x3d7,-1,-8},
3439 	{0x3d9,0x3ef,2,-1},
3440 	{0x3f0,0x3f0,-1,-86},
3441 	{0x3f1,0x3f1,-1,-80},
3442 	{0x3f2,0x3f2,-1,7},
3443 	{0x3f3,0x3f3,-1,-116},
3444 	{0x3f5,0x3f5,-1,-96},
3445 	{0x3f8,0x3fb,3,-1},
3446 	{0x430,0x44f,1,-32},
3447 	{0x450,0x45f,1,-80},
3448 	{0x461,0x481,2,-1},
3449 	{0x48b,0x4bf,2,-1},
3450 	{0x4c2,0x4ce,2,-1},
3451 	{0x4cf,0x4cf,-1,-15},
3452 	{0x4d1,0x52f,2,-1},
3453 	{0x561,0x586,1,-48},
3454 	{0x10d0,0x10fa,1,3008},
3455 	{0x10fd,0x10ff,1,3008},
3456 	{0x13f8,0x13fd,1,-8},
3457 	{0x1c80,0x1c80,-1,-6254},
3458 	{0x1c81,0x1c81,-1,-6253},
3459 	{0x1c82,0x1c82,-1,-6244},
3460 	{0x1c83,0x1c84,1,-6242},
3461 	{0x1c85,0x1c85,-1,-6243},
3462 	{0x1c86,0x1c86,-1,-6236},
3463 	{0x1c87,0x1c87,-1,-6181},
3464 	{0x1c88,0x1c88,-1,35266},
3465 	{0x1d79,0x1d79,-1,35332},
3466 	{0x1d7d,0x1d7d,-1,3814},
3467 	{0x1d8e,0x1d8e,-1,35384},
3468 	{0x1e01,0x1e95,2,-1},
3469 	{0x1e9b,0x1e9b,-1,-59},
3470 	{0x1ea1,0x1eff,2,-1},
3471 	{0x1f00,0x1f07,1,8},
3472 	{0x1f10,0x1f15,1,8},
3473 	{0x1f20,0x1f27,1,8},
3474 	{0x1f30,0x1f37,1,8},
3475 	{0x1f40,0x1f45,1,8},
3476 	{0x1f51,0x1f57,2,8},
3477 	{0x1f60,0x1f67,1,8},
3478 	{0x1f70,0x1f71,1,74},
3479 	{0x1f72,0x1f75,1,86},
3480 	{0x1f76,0x1f77,1,100},
3481 	{0x1f78,0x1f79,1,128},
3482 	{0x1f7a,0x1f7b,1,112},
3483 	{0x1f7c,0x1f7d,1,126},
3484 	{0x1f80,0x1f87,1,8},
3485 	{0x1f90,0x1f97,1,8},
3486 	{0x1fa0,0x1fa7,1,8},
3487 	{0x1fb0,0x1fb1,1,8},
3488 	{0x1fb3,0x1fb3,-1,9},
3489 	{0x1fbe,0x1fbe,-1,-7205},
3490 	{0x1fc3,0x1fc3,-1,9},
3491 	{0x1fd0,0x1fd1,1,8},
3492 	{0x1fe0,0x1fe1,1,8},
3493 	{0x1fe5,0x1fe5,-1,7},
3494 	{0x1ff3,0x1ff3,-1,9},
3495 	{0x214e,0x214e,-1,-28},
3496 	{0x2170,0x217f,1,-16},
3497 	{0x2184,0x2184,-1,-1},
3498 	{0x24d0,0x24e9,1,-26},
3499 	{0x2c30,0x2c5e,1,-48},
3500 	{0x2c61,0x2c61,-1,-1},
3501 	{0x2c65,0x2c65,-1,-10795},
3502 	{0x2c66,0x2c66,-1,-10792},
3503 	{0x2c68,0x2c6c,2,-1},
3504 	{0x2c73,0x2c76,3,-1},
3505 	{0x2c81,0x2ce3,2,-1},
3506 	{0x2cec,0x2cee,2,-1},
3507 	{0x2cf3,0x2cf3,-1,-1},
3508 	{0x2d00,0x2d25,1,-7264},
3509 	{0x2d27,0x2d2d,6,-7264},
3510 	{0xa641,0xa66d,2,-1},
3511 	{0xa681,0xa69b,2,-1},
3512 	{0xa723,0xa72f,2,-1},
3513 	{0xa733,0xa76f,2,-1},
3514 	{0xa77a,0xa77c,2,-1},
3515 	{0xa77f,0xa787,2,-1},
3516 	{0xa78c,0xa791,5,-1},
3517 	{0xa793,0xa793,-1,-1},
3518 	{0xa794,0xa794,-1,48},
3519 	{0xa797,0xa7a9,2,-1},
3520 	{0xa7b5,0xa7bf,2,-1},
3521 	{0xa7c3,0xa7c3,-1,-1},
3522 	{0xab53,0xab53,-1,-928},
3523 	{0xab70,0xabbf,1,-38864},
3524 	{0xff41,0xff5a,1,-32},
3525 	{0x10428,0x1044f,1,-40},
3526 	{0x104d8,0x104fb,1,-40},
3527 	{0x10cc0,0x10cf2,1,-64},
3528 	{0x118c0,0x118df,1,-32},
3529 	{0x16e60,0x16e7f,1,-32},
3530 	{0x1e922,0x1e943,1,-34}
3531 };
3532 
3533 /*
3534  * Return the upper-case equivalent of "a", which is a UCS-4 character.  Use
3535  * simple case folding.
3536  */
3537     int
3538 utf_toupper(int a)
3539 {
3540     /* If 'casemap' contains "keepascii" use ASCII style toupper(). */
3541     if (a < 128 && (cmp_flags & CMP_KEEPASCII))
3542 	return TOUPPER_ASC(a);
3543 
3544 #if defined(HAVE_TOWUPPER) && defined(__STDC_ISO_10646__)
3545     /* If towupper() is available and handles Unicode, use it. */
3546     if (!(cmp_flags & CMP_INTERNAL))
3547 	return towupper(a);
3548 #endif
3549 
3550     /* For characters below 128 use locale sensitive toupper(). */
3551     if (a < 128)
3552 	return TOUPPER_LOC(a);
3553 
3554     /* For any other characters use the above mapping table. */
3555     return utf_convert(a, toUpper, (int)sizeof(toUpper));
3556 }
3557 
3558     int
3559 utf_islower(int a)
3560 {
3561     /* German sharp s is lower case but has no upper case equivalent. */
3562     return (utf_toupper(a) != a) || a == 0xdf;
3563 }
3564 
3565 /*
3566  * Return the lower-case equivalent of "a", which is a UCS-4 character.  Use
3567  * simple case folding.
3568  */
3569     int
3570 utf_tolower(int a)
3571 {
3572     /* If 'casemap' contains "keepascii" use ASCII style tolower(). */
3573     if (a < 128 && (cmp_flags & CMP_KEEPASCII))
3574 	return TOLOWER_ASC(a);
3575 
3576 #if defined(HAVE_TOWLOWER) && defined(__STDC_ISO_10646__)
3577     /* If towlower() is available and handles Unicode, use it. */
3578     if (!(cmp_flags & CMP_INTERNAL))
3579 	return towlower(a);
3580 #endif
3581 
3582     /* For characters below 128 use locale sensitive tolower(). */
3583     if (a < 128)
3584 	return TOLOWER_LOC(a);
3585 
3586     /* For any other characters use the above mapping table. */
3587     return utf_convert(a, toLower, (int)sizeof(toLower));
3588 }
3589 
3590     int
3591 utf_isupper(int a)
3592 {
3593     return (utf_tolower(a) != a);
3594 }
3595 
3596     static int
3597 utf_strnicmp(
3598     char_u      *s1,
3599     char_u      *s2,
3600     size_t      n1,
3601     size_t      n2)
3602 {
3603     int		c1, c2, cdiff;
3604     char_u	buffer[6];
3605 
3606     for (;;)
3607     {
3608 	c1 = utf_safe_read_char_adv(&s1, &n1);
3609 	c2 = utf_safe_read_char_adv(&s2, &n2);
3610 
3611 	if (c1 <= 0 || c2 <= 0)
3612 	    break;
3613 
3614 	if (c1 == c2)
3615 	    continue;
3616 
3617 	cdiff = utf_fold(c1) - utf_fold(c2);
3618 	if (cdiff != 0)
3619 	    return cdiff;
3620     }
3621 
3622     /* some string ended or has an incomplete/illegal character sequence */
3623 
3624     if (c1 == 0 || c2 == 0)
3625     {
3626 	/* some string ended. shorter string is smaller */
3627 	if (c1 == 0 && c2 == 0)
3628 	    return 0;
3629 	return c1 == 0 ? -1 : 1;
3630     }
3631 
3632     /* Continue with bytewise comparison to produce some result that
3633      * would make comparison operations involving this function transitive.
3634      *
3635      * If only one string had an error, comparison should be made with
3636      * folded version of the other string. In this case it is enough
3637      * to fold just one character to determine the result of comparison. */
3638 
3639     if (c1 != -1 && c2 == -1)
3640     {
3641 	n1 = utf_char2bytes(utf_fold(c1), buffer);
3642 	s1 = buffer;
3643     }
3644     else if (c2 != -1 && c1 == -1)
3645     {
3646 	n2 = utf_char2bytes(utf_fold(c2), buffer);
3647 	s2 = buffer;
3648     }
3649 
3650     while (n1 > 0 && n2 > 0 && *s1 != NUL && *s2 != NUL)
3651     {
3652 	cdiff = (int)(*s1) - (int)(*s2);
3653 	if (cdiff != 0)
3654 	    return cdiff;
3655 
3656 	s1++;
3657 	s2++;
3658 	n1--;
3659 	n2--;
3660     }
3661 
3662     if (n1 > 0 && *s1 == NUL)
3663 	n1 = 0;
3664     if (n2 > 0 && *s2 == NUL)
3665 	n2 = 0;
3666 
3667     if (n1 == 0 && n2 == 0)
3668 	return 0;
3669     return n1 == 0 ? -1 : 1;
3670 }
3671 
3672 /*
3673  * Version of strnicmp() that handles multi-byte characters.
3674  * Needed for Big5, Shift-JIS and UTF-8 encoding.  Other DBCS encodings can
3675  * probably use strnicmp(), because there are no ASCII characters in the
3676  * second byte.
3677  * Returns zero if s1 and s2 are equal (ignoring case), the difference between
3678  * two characters otherwise.
3679  */
3680     int
3681 mb_strnicmp(char_u *s1, char_u *s2, size_t nn)
3682 {
3683     int		i, l;
3684     int		cdiff;
3685     int		n = (int)nn;
3686 
3687     if (enc_utf8)
3688     {
3689 	return utf_strnicmp(s1, s2, nn, nn);
3690     }
3691     else
3692     {
3693 	for (i = 0; i < n; i += l)
3694 	{
3695 	    if (s1[i] == NUL && s2[i] == NUL)	/* both strings end */
3696 		return 0;
3697 
3698 	    l = (*mb_ptr2len)(s1 + i);
3699 	    if (l <= 1)
3700 	    {
3701 		/* Single byte: first check normally, then with ignore case. */
3702 		if (s1[i] != s2[i])
3703 		{
3704 		    cdiff = MB_TOLOWER(s1[i]) - MB_TOLOWER(s2[i]);
3705 		    if (cdiff != 0)
3706 			return cdiff;
3707 		}
3708 	    }
3709 	    else
3710 	    {
3711 		/* For non-Unicode multi-byte don't ignore case. */
3712 		if (l > n - i)
3713 		    l = n - i;
3714 		cdiff = STRNCMP(s1 + i, s2 + i, l);
3715 		if (cdiff != 0)
3716 		    return cdiff;
3717 	    }
3718 	}
3719     }
3720     return 0;
3721 }
3722 
3723 /*
3724  * "g8": show bytes of the UTF-8 char under the cursor.  Doesn't matter what
3725  * 'encoding' has been set to.
3726  */
3727     void
3728 show_utf8(void)
3729 {
3730     int		len;
3731     int		rlen = 0;
3732     char_u	*line;
3733     int		clen;
3734     int		i;
3735 
3736     /* Get the byte length of the char under the cursor, including composing
3737      * characters. */
3738     line = ml_get_cursor();
3739     len = utfc_ptr2len(line);
3740     if (len == 0)
3741     {
3742 	msg("NUL");
3743 	return;
3744     }
3745 
3746     clen = 0;
3747     for (i = 0; i < len; ++i)
3748     {
3749 	if (clen == 0)
3750 	{
3751 	    /* start of (composing) character, get its length */
3752 	    if (i > 0)
3753 	    {
3754 		STRCPY(IObuff + rlen, "+ ");
3755 		rlen += 2;
3756 	    }
3757 	    clen = utf_ptr2len(line + i);
3758 	}
3759 	sprintf((char *)IObuff + rlen, "%02x ",
3760 		(line[i] == NL) ? NUL : line[i]);  /* NUL is stored as NL */
3761 	--clen;
3762 	rlen += (int)STRLEN(IObuff + rlen);
3763 	if (rlen > IOSIZE - 20)
3764 	    break;
3765     }
3766 
3767     msg((char *)IObuff);
3768 }
3769 
3770 /*
3771  * mb_head_off() function pointer.
3772  * Return offset from "p" to the first byte of the character it points into.
3773  * If "p" points to the NUL at the end of the string return 0.
3774  * Returns 0 when already at the first byte of a character.
3775  */
3776     int
3777 latin_head_off(char_u *base UNUSED, char_u *p UNUSED)
3778 {
3779     return 0;
3780 }
3781 
3782     int
3783 dbcs_head_off(char_u *base, char_u *p)
3784 {
3785     char_u	*q;
3786 
3787     /* It can't be a trailing byte when not using DBCS, at the start of the
3788      * string or the previous byte can't start a double-byte. */
3789     if (p <= base || MB_BYTE2LEN(p[-1]) == 1 || *p == NUL)
3790 	return 0;
3791 
3792     /* This is slow: need to start at the base and go forward until the
3793      * byte we are looking for.  Return 1 when we went past it, 0 otherwise. */
3794     q = base;
3795     while (q < p)
3796 	q += dbcs_ptr2len(q);
3797     return (q == p) ? 0 : 1;
3798 }
3799 
3800 /*
3801  * Special version of dbcs_head_off() that works for ScreenLines[], where
3802  * single-width DBCS_JPNU characters are stored separately.
3803  */
3804     int
3805 dbcs_screen_head_off(char_u *base, char_u *p)
3806 {
3807     char_u	*q;
3808 
3809     /* It can't be a trailing byte when not using DBCS, at the start of the
3810      * string or the previous byte can't start a double-byte.
3811      * For euc-jp an 0x8e byte in the previous cell always means we have a
3812      * lead byte in the current cell. */
3813     if (p <= base
3814 	    || (enc_dbcs == DBCS_JPNU && p[-1] == 0x8e)
3815 	    || MB_BYTE2LEN(p[-1]) == 1
3816 	    || *p == NUL)
3817 	return 0;
3818 
3819     /* This is slow: need to start at the base and go forward until the
3820      * byte we are looking for.  Return 1 when we went past it, 0 otherwise.
3821      * For DBCS_JPNU look out for 0x8e, which means the second byte is not
3822      * stored as the next byte. */
3823     q = base;
3824     while (q < p)
3825     {
3826 	if (enc_dbcs == DBCS_JPNU && *q == 0x8e)
3827 	    ++q;
3828 	else
3829 	    q += dbcs_ptr2len(q);
3830     }
3831     return (q == p) ? 0 : 1;
3832 }
3833 
3834     int
3835 utf_head_off(char_u *base, char_u *p)
3836 {
3837     char_u	*q;
3838     char_u	*s;
3839     int		c;
3840     int		len;
3841 #ifdef FEAT_ARABIC
3842     char_u	*j;
3843 #endif
3844 
3845     if (*p < 0x80)		/* be quick for ASCII */
3846 	return 0;
3847 
3848     /* Skip backwards over trailing bytes: 10xx.xxxx
3849      * Skip backwards again if on a composing char. */
3850     for (q = p; ; --q)
3851     {
3852 	/* Move s to the last byte of this char. */
3853 	for (s = q; (s[1] & 0xc0) == 0x80; ++s)
3854 	    ;
3855 	/* Move q to the first byte of this char. */
3856 	while (q > base && (*q & 0xc0) == 0x80)
3857 	    --q;
3858 	/* Check for illegal sequence. Do allow an illegal byte after where we
3859 	 * started. */
3860 	len = utf8len_tab[*q];
3861 	if (len != (int)(s - q + 1) && len != (int)(p - q + 1))
3862 	    return 0;
3863 
3864 	if (q <= base)
3865 	    break;
3866 
3867 	c = utf_ptr2char(q);
3868 	if (utf_iscomposing(c))
3869 	    continue;
3870 
3871 #ifdef FEAT_ARABIC
3872 	if (arabic_maycombine(c))
3873 	{
3874 	    /* Advance to get a sneak-peak at the next char */
3875 	    j = q;
3876 	    --j;
3877 	    /* Move j to the first byte of this char. */
3878 	    while (j > base && (*j & 0xc0) == 0x80)
3879 		--j;
3880 	    if (arabic_combine(utf_ptr2char(j), c))
3881 		continue;
3882 	}
3883 #endif
3884 	break;
3885     }
3886 
3887     return (int)(p - q);
3888 }
3889 
3890 /*
3891  * Copy a character from "*fp" to "*tp" and advance the pointers.
3892  */
3893     void
3894 mb_copy_char(char_u **fp, char_u **tp)
3895 {
3896     int	    l = (*mb_ptr2len)(*fp);
3897 
3898     mch_memmove(*tp, *fp, (size_t)l);
3899     *tp += l;
3900     *fp += l;
3901 }
3902 
3903 /*
3904  * Return the offset from "p" to the first byte of a character.  When "p" is
3905  * at the start of a character 0 is returned, otherwise the offset to the next
3906  * character.  Can start anywhere in a stream of bytes.
3907  */
3908     int
3909 mb_off_next(char_u *base, char_u *p)
3910 {
3911     int		i;
3912     int		j;
3913 
3914     if (enc_utf8)
3915     {
3916 	if (*p < 0x80)		/* be quick for ASCII */
3917 	    return 0;
3918 
3919 	/* Find the next character that isn't 10xx.xxxx */
3920 	for (i = 0; (p[i] & 0xc0) == 0x80; ++i)
3921 	    ;
3922 	if (i > 0)
3923 	{
3924 	    /* Check for illegal sequence. */
3925 	    for (j = 0; p - j > base; ++j)
3926 		if ((p[-j] & 0xc0) != 0x80)
3927 		    break;
3928 	    if (utf8len_tab[p[-j]] != i + j)
3929 		return 0;
3930 	}
3931 	return i;
3932     }
3933 
3934     /* Only need to check if we're on a trail byte, it doesn't matter if we
3935      * want the offset to the next or current character. */
3936     return (*mb_head_off)(base, p);
3937 }
3938 
3939 /*
3940  * Return the offset from "p" to the last byte of the character it points
3941  * into.  Can start anywhere in a stream of bytes.
3942  */
3943     int
3944 mb_tail_off(char_u *base, char_u *p)
3945 {
3946     int		i;
3947     int		j;
3948 
3949     if (*p == NUL)
3950 	return 0;
3951 
3952     if (enc_utf8)
3953     {
3954 	/* Find the last character that is 10xx.xxxx */
3955 	for (i = 0; (p[i + 1] & 0xc0) == 0x80; ++i)
3956 	    ;
3957 	/* Check for illegal sequence. */
3958 	for (j = 0; p - j > base; ++j)
3959 	    if ((p[-j] & 0xc0) != 0x80)
3960 		break;
3961 	if (utf8len_tab[p[-j]] != i + j + 1)
3962 	    return 0;
3963 	return i;
3964     }
3965 
3966     /* It can't be the first byte if a double-byte when not using DBCS, at the
3967      * end of the string or the byte can't start a double-byte. */
3968     if (enc_dbcs == 0 || p[1] == NUL || MB_BYTE2LEN(*p) == 1)
3969 	return 0;
3970 
3971     /* Return 1 when on the lead byte, 0 when on the tail byte. */
3972     return 1 - dbcs_head_off(base, p);
3973 }
3974 
3975 /*
3976  * Find the next illegal byte sequence.
3977  */
3978     void
3979 utf_find_illegal(void)
3980 {
3981     pos_T	pos = curwin->w_cursor;
3982     char_u	*p;
3983     int		len;
3984     vimconv_T	vimconv;
3985     char_u	*tofree = NULL;
3986 
3987     vimconv.vc_type = CONV_NONE;
3988     if (enc_utf8 && (enc_canon_props(curbuf->b_p_fenc) & ENC_8BIT))
3989     {
3990 	/* 'encoding' is "utf-8" but we are editing a 8-bit encoded file,
3991 	 * possibly a utf-8 file with illegal bytes.  Setup for conversion
3992 	 * from utf-8 to 'fileencoding'. */
3993 	convert_setup(&vimconv, p_enc, curbuf->b_p_fenc);
3994     }
3995 
3996     curwin->w_cursor.coladd = 0;
3997     for (;;)
3998     {
3999 	p = ml_get_cursor();
4000 	if (vimconv.vc_type != CONV_NONE)
4001 	{
4002 	    vim_free(tofree);
4003 	    tofree = string_convert(&vimconv, p, NULL);
4004 	    if (tofree == NULL)
4005 		break;
4006 	    p = tofree;
4007 	}
4008 
4009 	while (*p != NUL)
4010 	{
4011 	    /* Illegal means that there are not enough trail bytes (checked by
4012 	     * utf_ptr2len()) or too many of them (overlong sequence). */
4013 	    len = utf_ptr2len(p);
4014 	    if (*p >= 0x80 && (len == 1
4015 				     || utf_char2len(utf_ptr2char(p)) != len))
4016 	    {
4017 		if (vimconv.vc_type == CONV_NONE)
4018 		    curwin->w_cursor.col += (colnr_T)(p - ml_get_cursor());
4019 		else
4020 		{
4021 		    int	    l;
4022 
4023 		    len = (int)(p - tofree);
4024 		    for (p = ml_get_cursor(); *p != NUL && len-- > 0; p += l)
4025 		    {
4026 			l = utf_ptr2len(p);
4027 			curwin->w_cursor.col += l;
4028 		    }
4029 		}
4030 		goto theend;
4031 	    }
4032 	    p += len;
4033 	}
4034 	if (curwin->w_cursor.lnum == curbuf->b_ml.ml_line_count)
4035 	    break;
4036 	++curwin->w_cursor.lnum;
4037 	curwin->w_cursor.col = 0;
4038     }
4039 
4040     /* didn't find it: don't move and beep */
4041     curwin->w_cursor = pos;
4042     beep_flush();
4043 
4044 theend:
4045     vim_free(tofree);
4046     convert_setup(&vimconv, NULL, NULL);
4047 }
4048 
4049 #if defined(FEAT_GUI_GTK) || defined(PROTO)
4050 /*
4051  * Return TRUE if string "s" is a valid utf-8 string.
4052  * When "end" is NULL stop at the first NUL.
4053  * When "end" is positive stop there.
4054  */
4055     int
4056 utf_valid_string(char_u *s, char_u *end)
4057 {
4058     int		l;
4059     char_u	*p = s;
4060 
4061     while (end == NULL ? *p != NUL : p < end)
4062     {
4063 	l = utf8len_tab_zero[*p];
4064 	if (l == 0)
4065 	    return FALSE;	/* invalid lead byte */
4066 	if (end != NULL && p + l > end)
4067 	    return FALSE;	/* incomplete byte sequence */
4068 	++p;
4069 	while (--l > 0)
4070 	    if ((*p++ & 0xc0) != 0x80)
4071 		return FALSE;	/* invalid trail byte */
4072     }
4073     return TRUE;
4074 }
4075 #endif
4076 
4077 #if defined(FEAT_GUI) || defined(PROTO)
4078 /*
4079  * Special version of mb_tail_off() for use in ScreenLines[].
4080  */
4081     int
4082 dbcs_screen_tail_off(char_u *base, char_u *p)
4083 {
4084     /* It can't be the first byte if a double-byte when not using DBCS, at the
4085      * end of the string or the byte can't start a double-byte.
4086      * For euc-jp an 0x8e byte always means we have a lead byte in the current
4087      * cell. */
4088     if (*p == NUL || p[1] == NUL
4089 	    || (enc_dbcs == DBCS_JPNU && *p == 0x8e)
4090 	    || MB_BYTE2LEN(*p) == 1)
4091 	return 0;
4092 
4093     /* Return 1 when on the lead byte, 0 when on the tail byte. */
4094     return 1 - dbcs_screen_head_off(base, p);
4095 }
4096 #endif
4097 
4098 /*
4099  * If the cursor moves on an trail byte, set the cursor on the lead byte.
4100  * Thus it moves left if necessary.
4101  * Return TRUE when the cursor was adjusted.
4102  */
4103     void
4104 mb_adjust_cursor(void)
4105 {
4106     mb_adjustpos(curbuf, &curwin->w_cursor);
4107 }
4108 
4109 /*
4110  * Adjust position "*lp" to point to the first byte of a multi-byte character.
4111  * If it points to a tail byte it's moved backwards to the head byte.
4112  */
4113     void
4114 mb_adjustpos(buf_T *buf, pos_T *lp)
4115 {
4116     char_u	*p;
4117 
4118     if (lp->col > 0 || lp->coladd > 1)
4119     {
4120 	p = ml_get_buf(buf, lp->lnum, FALSE);
4121 	if (*p == NUL || (int)STRLEN(p) < lp->col)
4122 	    lp->col = 0;
4123 	else
4124 	    lp->col -= (*mb_head_off)(p, p + lp->col);
4125 	/* Reset "coladd" when the cursor would be on the right half of a
4126 	 * double-wide character. */
4127 	if (lp->coladd == 1
4128 		&& p[lp->col] != TAB
4129 		&& vim_isprintc((*mb_ptr2char)(p + lp->col))
4130 		&& ptr2cells(p + lp->col) > 1)
4131 	    lp->coladd = 0;
4132     }
4133 }
4134 
4135 /*
4136  * Return a pointer to the character before "*p", if there is one.
4137  */
4138     char_u *
4139 mb_prevptr(
4140     char_u *line,	/* start of the string */
4141     char_u *p)
4142 {
4143     if (p > line)
4144 	MB_PTR_BACK(line, p);
4145     return p;
4146 }
4147 
4148 /*
4149  * Return the character length of "str".  Each multi-byte character (with
4150  * following composing characters) counts as one.
4151  */
4152     int
4153 mb_charlen(char_u *str)
4154 {
4155     char_u	*p = str;
4156     int		count;
4157 
4158     if (p == NULL)
4159 	return 0;
4160 
4161     for (count = 0; *p != NUL; count++)
4162 	p += (*mb_ptr2len)(p);
4163 
4164     return count;
4165 }
4166 
4167 #if defined(FEAT_SPELL) || defined(PROTO)
4168 /*
4169  * Like mb_charlen() but for a string with specified length.
4170  */
4171     int
4172 mb_charlen_len(char_u *str, int len)
4173 {
4174     char_u	*p = str;
4175     int		count;
4176 
4177     for (count = 0; *p != NUL && p < str + len; count++)
4178 	p += (*mb_ptr2len)(p);
4179 
4180     return count;
4181 }
4182 #endif
4183 
4184 /*
4185  * Try to un-escape a multi-byte character.
4186  * Used for the "to" and "from" part of a mapping.
4187  * Return the un-escaped string if it is a multi-byte character, and advance
4188  * "pp" to just after the bytes that formed it.
4189  * Return NULL if no multi-byte char was found.
4190  */
4191     char_u *
4192 mb_unescape(char_u **pp)
4193 {
4194     static char_u	buf[6];
4195     int			n;
4196     int			m = 0;
4197     char_u		*str = *pp;
4198 
4199     /* Must translate K_SPECIAL KS_SPECIAL KE_FILLER to K_SPECIAL and CSI
4200      * KS_EXTRA KE_CSI to CSI.
4201      * Maximum length of a utf-8 character is 4 bytes. */
4202     for (n = 0; str[n] != NUL && m < 4; ++n)
4203     {
4204 	if (str[n] == K_SPECIAL
4205 		&& str[n + 1] == KS_SPECIAL
4206 		&& str[n + 2] == KE_FILLER)
4207 	{
4208 	    buf[m++] = K_SPECIAL;
4209 	    n += 2;
4210 	}
4211 	else if ((str[n] == K_SPECIAL
4212 # ifdef FEAT_GUI
4213 		    || str[n] == CSI
4214 # endif
4215 		 )
4216 		&& str[n + 1] == KS_EXTRA
4217 		&& str[n + 2] == (int)KE_CSI)
4218 	{
4219 	    buf[m++] = CSI;
4220 	    n += 2;
4221 	}
4222 	else if (str[n] == K_SPECIAL
4223 # ifdef FEAT_GUI
4224 		|| str[n] == CSI
4225 # endif
4226 		)
4227 	    break;		/* a special key can't be a multibyte char */
4228 	else
4229 	    buf[m++] = str[n];
4230 	buf[m] = NUL;
4231 
4232 	/* Return a multi-byte character if it's found.  An illegal sequence
4233 	 * will result in a 1 here. */
4234 	if ((*mb_ptr2len)(buf) > 1)
4235 	{
4236 	    *pp = str + n + 1;
4237 	    return buf;
4238 	}
4239 
4240 	/* Bail out quickly for ASCII. */
4241 	if (buf[0] < 128)
4242 	    break;
4243     }
4244     return NULL;
4245 }
4246 
4247 /*
4248  * Return TRUE if the character at "row"/"col" on the screen is the left side
4249  * of a double-width character.
4250  * Caller must make sure "row" and "col" are not invalid!
4251  */
4252     int
4253 mb_lefthalve(int row, int col)
4254 {
4255 #ifdef FEAT_HANGULIN
4256     if (composing_hangul)
4257 	return TRUE;
4258 #endif
4259     return (*mb_off2cells)(LineOffset[row] + col,
4260 					LineOffset[row] + screen_Columns) > 1;
4261 }
4262 
4263 /*
4264  * Correct a position on the screen, if it's the right half of a double-wide
4265  * char move it to the left half.  Returns the corrected column.
4266  */
4267     int
4268 mb_fix_col(int col, int row)
4269 {
4270     col = check_col(col);
4271     row = check_row(row);
4272     if (has_mbyte && ScreenLines != NULL && col > 0
4273 	    && ((enc_dbcs
4274 		    && ScreenLines[LineOffset[row] + col] != NUL
4275 		    && dbcs_screen_head_off(ScreenLines + LineOffset[row],
4276 					 ScreenLines + LineOffset[row] + col))
4277 		|| (enc_utf8 && ScreenLines[LineOffset[row] + col] == 0)))
4278 	return col - 1;
4279     return col;
4280 }
4281 
4282 static int enc_alias_search(char_u *name);
4283 
4284 /*
4285  * Skip the Vim specific head of a 'encoding' name.
4286  */
4287     char_u *
4288 enc_skip(char_u *p)
4289 {
4290     if (STRNCMP(p, "2byte-", 6) == 0)
4291 	return p + 6;
4292     if (STRNCMP(p, "8bit-", 5) == 0)
4293 	return p + 5;
4294     return p;
4295 }
4296 
4297 /*
4298  * Find the canonical name for encoding "enc".
4299  * When the name isn't recognized, returns "enc" itself, but with all lower
4300  * case characters and '_' replaced with '-'.
4301  * Returns an allocated string.  NULL for out-of-memory.
4302  */
4303     char_u *
4304 enc_canonize(char_u *enc)
4305 {
4306     char_u	*r;
4307     char_u	*p, *s;
4308     int		i;
4309 
4310     if (STRCMP(enc, "default") == 0)
4311     {
4312 	/* Use the default encoding as it's found by set_init_1(). */
4313 	r = get_encoding_default();
4314 	if (r == NULL)
4315 	    r = (char_u *)"latin1";
4316 	return vim_strsave(r);
4317     }
4318 
4319     /* copy "enc" to allocated memory, with room for two '-' */
4320     r = alloc((unsigned)(STRLEN(enc) + 3));
4321     if (r != NULL)
4322     {
4323 	/* Make it all lower case and replace '_' with '-'. */
4324 	p = r;
4325 	for (s = enc; *s != NUL; ++s)
4326 	{
4327 	    if (*s == '_')
4328 		*p++ = '-';
4329 	    else
4330 		*p++ = TOLOWER_ASC(*s);
4331 	}
4332 	*p = NUL;
4333 
4334 	/* Skip "2byte-" and "8bit-". */
4335 	p = enc_skip(r);
4336 
4337 	/* Change "microsoft-cp" to "cp".  Used in some spell files. */
4338 	if (STRNCMP(p, "microsoft-cp", 12) == 0)
4339 	    STRMOVE(p, p + 10);
4340 
4341 	/* "iso8859" -> "iso-8859" */
4342 	if (STRNCMP(p, "iso8859", 7) == 0)
4343 	{
4344 	    STRMOVE(p + 4, p + 3);
4345 	    p[3] = '-';
4346 	}
4347 
4348 	/* "iso-8859n" -> "iso-8859-n" */
4349 	if (STRNCMP(p, "iso-8859", 8) == 0 && p[8] != '-')
4350 	{
4351 	    STRMOVE(p + 9, p + 8);
4352 	    p[8] = '-';
4353 	}
4354 
4355 	/* "latin-N" -> "latinN" */
4356 	if (STRNCMP(p, "latin-", 6) == 0)
4357 	    STRMOVE(p + 5, p + 6);
4358 
4359 	if (enc_canon_search(p) >= 0)
4360 	{
4361 	    /* canonical name can be used unmodified */
4362 	    if (p != r)
4363 		STRMOVE(r, p);
4364 	}
4365 	else if ((i = enc_alias_search(p)) >= 0)
4366 	{
4367 	    /* alias recognized, get canonical name */
4368 	    vim_free(r);
4369 	    r = vim_strsave((char_u *)enc_canon_table[i].name);
4370 	}
4371     }
4372     return r;
4373 }
4374 
4375 /*
4376  * Search for an encoding alias of "name".
4377  * Returns -1 when not found.
4378  */
4379     static int
4380 enc_alias_search(char_u *name)
4381 {
4382     int		i;
4383 
4384     for (i = 0; enc_alias_table[i].name != NULL; ++i)
4385 	if (STRCMP(name, enc_alias_table[i].name) == 0)
4386 	    return enc_alias_table[i].canon;
4387     return -1;
4388 }
4389 
4390 
4391 #ifdef HAVE_LANGINFO_H
4392 # include <langinfo.h>
4393 #endif
4394 
4395 #if !defined(FEAT_GUI_MSWIN) || defined(VIMDLL)
4396 /*
4397  * Get the canonicalized encoding from the specified locale string "locale"
4398  * or from the environment variables LC_ALL, LC_CTYPE and LANG.
4399  * Returns an allocated string when successful, NULL when not.
4400  */
4401     char_u *
4402 enc_locale_env(char *locale)
4403 {
4404     char	*s = locale;
4405     char	*p;
4406     int		i;
4407     char	buf[50];
4408 
4409     if (s == NULL || *s == NUL)
4410 	if ((s = getenv("LC_ALL")) == NULL || *s == NUL)
4411 	    if ((s = getenv("LC_CTYPE")) == NULL || *s == NUL)
4412 		s = getenv("LANG");
4413 
4414     if (s == NULL || *s == NUL)
4415 	return NULL;
4416 
4417     /* The most generic locale format is:
4418      * language[_territory][.codeset][@modifier][+special][,[sponsor][_revision]]
4419      * If there is a '.' remove the part before it.
4420      * if there is something after the codeset, remove it.
4421      * Make the name lowercase and replace '_' with '-'.
4422      * Exception: "ja_JP.EUC" == "euc-jp", "zh_CN.EUC" = "euc-cn",
4423      * "ko_KR.EUC" == "euc-kr"
4424      */
4425     if ((p = (char *)vim_strchr((char_u *)s, '.')) != NULL)
4426     {
4427 	if (p > s + 2 && STRNICMP(p + 1, "EUC", 3) == 0
4428 			&& !isalnum((int)p[4]) && p[4] != '-' && p[-3] == '_')
4429 	{
4430 	    /* copy "XY.EUC" to "euc-XY" to buf[10] */
4431 	    STRCPY(buf + 10, "euc-");
4432 	    buf[14] = p[-2];
4433 	    buf[15] = p[-1];
4434 	    buf[16] = 0;
4435 	    s = buf + 10;
4436 	}
4437 	else
4438 	    s = p + 1;
4439     }
4440     for (i = 0; i < (int)sizeof(buf) - 1 && s[i] != NUL; ++i)
4441     {
4442 	if (s[i] == '_' || s[i] == '-')
4443 	    buf[i] = '-';
4444 	else if (isalnum((int)s[i]))
4445 	    buf[i] = TOLOWER_ASC(s[i]);
4446 	else
4447 	    break;
4448     }
4449     buf[i] = NUL;
4450 
4451     return enc_canonize((char_u *)buf);
4452 }
4453 #endif
4454 
4455 /*
4456  * Get the canonicalized encoding of the current locale.
4457  * Returns an allocated string when successful, NULL when not.
4458  */
4459     char_u *
4460 enc_locale(void)
4461 {
4462 #ifdef MSWIN
4463     char	buf[50];
4464     long	acp = GetACP();
4465 
4466     if (acp == 1200)
4467 	STRCPY(buf, "ucs-2le");
4468     else if (acp == 1252)	    /* cp1252 is used as latin1 */
4469 	STRCPY(buf, "latin1");
4470     else
4471 	sprintf(buf, "cp%ld", acp);
4472 
4473     return enc_canonize((char_u *)buf);
4474 #else
4475     char	*s;
4476 
4477 # ifdef HAVE_NL_LANGINFO_CODESET
4478     if ((s = nl_langinfo(CODESET)) == NULL || *s == NUL)
4479 # endif
4480 # if defined(HAVE_LOCALE_H) || defined(X_LOCALE)
4481 	if ((s = setlocale(LC_CTYPE, NULL)) == NULL || *s == NUL)
4482 # endif
4483 	    s = NULL;
4484 
4485     return enc_locale_env(s);
4486 #endif
4487 }
4488 
4489 # if defined(MSWIN) || defined(PROTO) || defined(FEAT_CYGWIN_WIN32_CLIPBOARD)
4490 /*
4491  * Convert an encoding name to an MS-Windows codepage.
4492  * Returns zero if no codepage can be figured out.
4493  */
4494     int
4495 encname2codepage(char_u *name)
4496 {
4497     int		cp;
4498     char_u	*p = name;
4499     int		idx;
4500 
4501     if (STRNCMP(p, "8bit-", 5) == 0)
4502 	p += 5;
4503     else if (STRNCMP(p_enc, "2byte-", 6) == 0)
4504 	p += 6;
4505 
4506     if (p[0] == 'c' && p[1] == 'p')
4507 	cp = atoi((char *)p + 2);
4508     else if ((idx = enc_canon_search(p)) >= 0)
4509 	cp = enc_canon_table[idx].codepage;
4510     else
4511 	return 0;
4512     if (IsValidCodePage(cp))
4513 	return cp;
4514     return 0;
4515 }
4516 # endif
4517 
4518 # if defined(USE_ICONV) || defined(PROTO)
4519 
4520 /*
4521  * Call iconv_open() with a check if iconv() works properly (there are broken
4522  * versions).
4523  * Returns (void *)-1 if failed.
4524  * (should return iconv_t, but that causes problems with prototypes).
4525  */
4526     void *
4527 my_iconv_open(char_u *to, char_u *from)
4528 {
4529     iconv_t	fd;
4530 #define ICONV_TESTLEN 400
4531     char_u	tobuf[ICONV_TESTLEN];
4532     char	*p;
4533     size_t	tolen;
4534     static int	iconv_ok = -1;
4535 
4536     if (iconv_ok == FALSE)
4537 	return (void *)-1;	/* detected a broken iconv() previously */
4538 
4539 #ifdef DYNAMIC_ICONV
4540     /* Check if the iconv.dll can be found. */
4541     if (!iconv_enabled(TRUE))
4542 	return (void *)-1;
4543 #endif
4544 
4545     fd = iconv_open((char *)enc_skip(to), (char *)enc_skip(from));
4546 
4547     if (fd != (iconv_t)-1 && iconv_ok == -1)
4548     {
4549 	/*
4550 	 * Do a dummy iconv() call to check if it actually works.  There is a
4551 	 * version of iconv() on Linux that is broken.  We can't ignore it,
4552 	 * because it's wide-spread.  The symptoms are that after outputting
4553 	 * the initial shift state the "to" pointer is NULL and conversion
4554 	 * stops for no apparent reason after about 8160 characters.
4555 	 */
4556 	p = (char *)tobuf;
4557 	tolen = ICONV_TESTLEN;
4558 	(void)iconv(fd, NULL, NULL, &p, &tolen);
4559 	if (p == NULL)
4560 	{
4561 	    iconv_ok = FALSE;
4562 	    iconv_close(fd);
4563 	    fd = (iconv_t)-1;
4564 	}
4565 	else
4566 	    iconv_ok = TRUE;
4567     }
4568 
4569     return (void *)fd;
4570 }
4571 
4572 /*
4573  * Convert the string "str[slen]" with iconv().
4574  * If "unconvlenp" is not NULL handle the string ending in an incomplete
4575  * sequence and set "*unconvlenp" to the length of it.
4576  * Returns the converted string in allocated memory.  NULL for an error.
4577  * If resultlenp is not NULL, sets it to the result length in bytes.
4578  */
4579     static char_u *
4580 iconv_string(
4581     vimconv_T	*vcp,
4582     char_u	*str,
4583     int		slen,
4584     int		*unconvlenp,
4585     int		*resultlenp)
4586 {
4587     const char	*from;
4588     size_t	fromlen;
4589     char	*to;
4590     size_t	tolen;
4591     size_t	len = 0;
4592     size_t	done = 0;
4593     char_u	*result = NULL;
4594     char_u	*p;
4595     int		l;
4596 
4597     from = (char *)str;
4598     fromlen = slen;
4599     for (;;)
4600     {
4601 	if (len == 0 || ICONV_ERRNO == ICONV_E2BIG)
4602 	{
4603 	    /* Allocate enough room for most conversions.  When re-allocating
4604 	     * increase the buffer size. */
4605 	    len = len + fromlen * 2 + 40;
4606 	    p = alloc((unsigned)len);
4607 	    if (p != NULL && done > 0)
4608 		mch_memmove(p, result, done);
4609 	    vim_free(result);
4610 	    result = p;
4611 	    if (result == NULL)	/* out of memory */
4612 		break;
4613 	}
4614 
4615 	to = (char *)result + done;
4616 	tolen = len - done - 2;
4617 	/* Avoid a warning for systems with a wrong iconv() prototype by
4618 	 * casting the second argument to void *. */
4619 	if (iconv(vcp->vc_fd, (void *)&from, &fromlen, &to, &tolen)
4620 								!= (size_t)-1)
4621 	{
4622 	    /* Finished, append a NUL. */
4623 	    *to = NUL;
4624 	    break;
4625 	}
4626 
4627 	/* Check both ICONV_EINVAL and EINVAL, because the dynamically loaded
4628 	 * iconv library may use one of them. */
4629 	if (!vcp->vc_fail && unconvlenp != NULL
4630 		&& (ICONV_ERRNO == ICONV_EINVAL || ICONV_ERRNO == EINVAL))
4631 	{
4632 	    /* Handle an incomplete sequence at the end. */
4633 	    *to = NUL;
4634 	    *unconvlenp = (int)fromlen;
4635 	    break;
4636 	}
4637 
4638 	/* Check both ICONV_EILSEQ and EILSEQ, because the dynamically loaded
4639 	 * iconv library may use one of them. */
4640 	else if (!vcp->vc_fail
4641 		&& (ICONV_ERRNO == ICONV_EILSEQ || ICONV_ERRNO == EILSEQ
4642 		    || ICONV_ERRNO == ICONV_EINVAL || ICONV_ERRNO == EINVAL))
4643 	{
4644 	    /* Can't convert: insert a '?' and skip a character.  This assumes
4645 	     * conversion from 'encoding' to something else.  In other
4646 	     * situations we don't know what to skip anyway. */
4647 	    *to++ = '?';
4648 	    if ((*mb_ptr2cells)((char_u *)from) > 1)
4649 		*to++ = '?';
4650 	    if (enc_utf8)
4651 		l = utfc_ptr2len_len((char_u *)from, (int)fromlen);
4652 	    else
4653 	    {
4654 		l = (*mb_ptr2len)((char_u *)from);
4655 		if (l > (int)fromlen)
4656 		    l = (int)fromlen;
4657 	    }
4658 	    from += l;
4659 	    fromlen -= l;
4660 	}
4661 	else if (ICONV_ERRNO != ICONV_E2BIG)
4662 	{
4663 	    /* conversion failed */
4664 	    VIM_CLEAR(result);
4665 	    break;
4666 	}
4667 	/* Not enough room or skipping illegal sequence. */
4668 	done = to - (char *)result;
4669     }
4670 
4671     if (resultlenp != NULL && result != NULL)
4672 	*resultlenp = (int)(to - (char *)result);
4673     return result;
4674 }
4675 
4676 #  if defined(DYNAMIC_ICONV) || defined(PROTO)
4677 /*
4678  * Dynamically load the "iconv.dll" on Win32.
4679  */
4680 
4681 #   ifndef DYNAMIC_ICONV	    /* must be generating prototypes */
4682 #    define HINSTANCE int
4683 #   endif
4684 static HINSTANCE hIconvDLL = 0;
4685 static HINSTANCE hMsvcrtDLL = 0;
4686 
4687 #   ifndef DYNAMIC_ICONV_DLL
4688 #    define DYNAMIC_ICONV_DLL "iconv.dll"
4689 #    define DYNAMIC_ICONV_DLL_ALT1 "libiconv.dll"
4690 #    define DYNAMIC_ICONV_DLL_ALT2 "libiconv2.dll"
4691 #    define DYNAMIC_ICONV_DLL_ALT3 "libiconv-2.dll"
4692 #   endif
4693 #   ifndef DYNAMIC_MSVCRT_DLL
4694 #    define DYNAMIC_MSVCRT_DLL "msvcrt.dll"
4695 #   endif
4696 
4697 /*
4698  * Try opening the iconv.dll and return TRUE if iconv() can be used.
4699  */
4700     int
4701 iconv_enabled(int verbose)
4702 {
4703     if (hIconvDLL != 0 && hMsvcrtDLL != 0)
4704 	return TRUE;
4705 
4706     /* The iconv DLL file goes under different names, try them all.
4707      * Do the "2" version first, it's newer. */
4708 #ifdef DYNAMIC_ICONV_DLL_ALT2
4709     if (hIconvDLL == 0)
4710 	hIconvDLL = vimLoadLib(DYNAMIC_ICONV_DLL_ALT2);
4711 #endif
4712 #ifdef DYNAMIC_ICONV_DLL_ALT3
4713     if (hIconvDLL == 0)
4714 	hIconvDLL = vimLoadLib(DYNAMIC_ICONV_DLL_ALT3);
4715 #endif
4716     if (hIconvDLL == 0)
4717 	hIconvDLL = vimLoadLib(DYNAMIC_ICONV_DLL);
4718 #ifdef DYNAMIC_ICONV_DLL_ALT1
4719     if (hIconvDLL == 0)
4720 	hIconvDLL = vimLoadLib(DYNAMIC_ICONV_DLL_ALT1);
4721 #endif
4722 
4723     if (hIconvDLL != 0)
4724 	hMsvcrtDLL = vimLoadLib(DYNAMIC_MSVCRT_DLL);
4725     if (hIconvDLL == 0 || hMsvcrtDLL == 0)
4726     {
4727 	/* Only give the message when 'verbose' is set, otherwise it might be
4728 	 * done whenever a conversion is attempted. */
4729 	if (verbose && p_verbose > 0)
4730 	{
4731 	    verbose_enter();
4732 	    semsg(_(e_loadlib),
4733 		    hIconvDLL == 0 ? DYNAMIC_ICONV_DLL : DYNAMIC_MSVCRT_DLL);
4734 	    verbose_leave();
4735 	}
4736 	iconv_end();
4737 	return FALSE;
4738     }
4739 
4740     iconv	= (void *)GetProcAddress(hIconvDLL, "libiconv");
4741     iconv_open	= (void *)GetProcAddress(hIconvDLL, "libiconv_open");
4742     iconv_close	= (void *)GetProcAddress(hIconvDLL, "libiconv_close");
4743     iconvctl	= (void *)GetProcAddress(hIconvDLL, "libiconvctl");
4744     iconv_errno	= get_dll_import_func(hIconvDLL, "_errno");
4745     if (iconv_errno == NULL)
4746 	iconv_errno = (void *)GetProcAddress(hMsvcrtDLL, "_errno");
4747     if (iconv == NULL || iconv_open == NULL || iconv_close == NULL
4748 	    || iconvctl == NULL || iconv_errno == NULL)
4749     {
4750 	iconv_end();
4751 	if (verbose && p_verbose > 0)
4752 	{
4753 	    verbose_enter();
4754 	    semsg(_(e_loadfunc), "for libiconv");
4755 	    verbose_leave();
4756 	}
4757 	return FALSE;
4758     }
4759     return TRUE;
4760 }
4761 
4762     void
4763 iconv_end(void)
4764 {
4765     /* Don't use iconv() when inputting or outputting characters. */
4766     if (input_conv.vc_type == CONV_ICONV)
4767 	convert_setup(&input_conv, NULL, NULL);
4768     if (output_conv.vc_type == CONV_ICONV)
4769 	convert_setup(&output_conv, NULL, NULL);
4770 
4771     if (hIconvDLL != 0)
4772 	FreeLibrary(hIconvDLL);
4773     if (hMsvcrtDLL != 0)
4774 	FreeLibrary(hMsvcrtDLL);
4775     hIconvDLL = 0;
4776     hMsvcrtDLL = 0;
4777 }
4778 #  endif /* DYNAMIC_ICONV */
4779 # endif /* USE_ICONV */
4780 
4781 
4782 #ifdef FEAT_GUI
4783 # define USE_IMACTIVATEFUNC (!gui.in_use && *p_imaf != NUL)
4784 # define USE_IMSTATUSFUNC (!gui.in_use && *p_imsf != NUL)
4785 #else
4786 # define USE_IMACTIVATEFUNC (*p_imaf != NUL)
4787 # define USE_IMSTATUSFUNC (*p_imsf != NUL)
4788 #endif
4789 
4790 #if defined(FEAT_EVAL) && (defined(FEAT_XIM) || defined(IME_WITHOUT_XIM))
4791     static void
4792 call_imactivatefunc(int active)
4793 {
4794     typval_T argv[2];
4795 
4796     argv[0].v_type = VAR_NUMBER;
4797     argv[0].vval.v_number = active ? 1 : 0;
4798     argv[1].v_type = VAR_UNKNOWN;
4799     (void)call_func_retnr(p_imaf, 1, argv);
4800 }
4801 
4802     static int
4803 call_imstatusfunc(void)
4804 {
4805     int is_active;
4806 
4807     /* FIXME: Don't execute user function in unsafe situation. */
4808     if (exiting || is_autocmd_blocked())
4809 	return FALSE;
4810     /* FIXME: :py print 'xxx' is shown duplicate result.
4811      * Use silent to avoid it. */
4812     ++msg_silent;
4813     is_active = call_func_retnr(p_imsf, 0, NULL);
4814     --msg_silent;
4815     return (is_active > 0);
4816 }
4817 #endif
4818 
4819 #if defined(FEAT_XIM) || defined(PROTO)
4820 
4821 # if defined(FEAT_GUI_GTK) || defined(PROTO)
4822 static int xim_has_preediting INIT(= FALSE);  /* IM current status */
4823 
4824 /*
4825  * Set preedit_start_col to the current cursor position.
4826  */
4827     static void
4828 init_preedit_start_col(void)
4829 {
4830     if (State & CMDLINE)
4831 	preedit_start_col = cmdline_getvcol_cursor();
4832     else if (curwin != NULL && curwin->w_buffer != NULL)
4833 	getvcol(curwin, &curwin->w_cursor, &preedit_start_col, NULL, NULL);
4834     /* Prevent that preediting marks the buffer as changed. */
4835     xim_changed_while_preediting = curbuf->b_changed;
4836 }
4837 
4838 static int im_is_active	       = FALSE;	/* IM is enabled for current mode    */
4839 static int preedit_is_active   = FALSE;
4840 static int im_preedit_cursor   = 0;	/* cursor offset in characters       */
4841 static int im_preedit_trailing = 0;	/* number of characters after cursor */
4842 
4843 static unsigned long im_commit_handler_id  = 0;
4844 static unsigned int  im_activatekey_keyval = GDK_VoidSymbol;
4845 static unsigned int  im_activatekey_state  = 0;
4846 
4847 static GtkWidget *preedit_window = NULL;
4848 static GtkWidget *preedit_label = NULL;
4849 
4850 static void im_preedit_window_set_position(void);
4851 
4852     void
4853 im_set_active(int active)
4854 {
4855     int was_active;
4856 
4857     was_active = !!im_get_status();
4858     im_is_active = (active && !p_imdisable);
4859 
4860     if (im_is_active != was_active)
4861 	xim_reset();
4862 }
4863 
4864     void
4865 xim_set_focus(int focus)
4866 {
4867     if (xic != NULL)
4868     {
4869 	if (focus)
4870 	    gtk_im_context_focus_in(xic);
4871 	else
4872 	    gtk_im_context_focus_out(xic);
4873     }
4874 }
4875 
4876     void
4877 im_set_position(int row, int col)
4878 {
4879     if (xic != NULL)
4880     {
4881 	GdkRectangle area;
4882 
4883 	area.x = FILL_X(col);
4884 	area.y = FILL_Y(row);
4885 	area.width  = gui.char_width * (mb_lefthalve(row, col) ? 2 : 1);
4886 	area.height = gui.char_height;
4887 
4888 	gtk_im_context_set_cursor_location(xic, &area);
4889 
4890 	if (p_imst == IM_OVER_THE_SPOT)
4891 	    im_preedit_window_set_position();
4892     }
4893 }
4894 
4895 #  if 0 || defined(PROTO) /* apparently only used in gui_x11.c */
4896     void
4897 xim_set_preedit(void)
4898 {
4899     im_set_position(gui.row, gui.col);
4900 }
4901 #  endif
4902 
4903     static void
4904 im_add_to_input(char_u *str, int len)
4905 {
4906     /* Convert from 'termencoding' (always "utf-8") to 'encoding' */
4907     if (input_conv.vc_type != CONV_NONE)
4908     {
4909 	str = string_convert(&input_conv, str, &len);
4910 	g_return_if_fail(str != NULL);
4911     }
4912 
4913     add_to_input_buf_csi(str, len);
4914 
4915     if (input_conv.vc_type != CONV_NONE)
4916 	vim_free(str);
4917 
4918     if (p_mh) /* blank out the pointer if necessary */
4919 	gui_mch_mousehide(TRUE);
4920 }
4921 
4922      static void
4923 im_preedit_window_set_position(void)
4924 {
4925     int x, y, width, height;
4926     int screen_x, screen_y, screen_width, screen_height;
4927 
4928     if (preedit_window == NULL)
4929 	return;
4930 
4931     gui_gtk_get_screen_geom_of_win(gui.drawarea,
4932 			  &screen_x, &screen_y, &screen_width, &screen_height);
4933     gdk_window_get_origin(gtk_widget_get_window(gui.drawarea), &x, &y);
4934     gtk_window_get_size(GTK_WINDOW(preedit_window), &width, &height);
4935     x = x + FILL_X(gui.col);
4936     y = y + FILL_Y(gui.row);
4937     if (x + width > screen_x + screen_width)
4938 	x = screen_x + screen_width - width;
4939     if (y + height > screen_y + screen_height)
4940 	y = screen_y + screen_height - height;
4941     gtk_window_move(GTK_WINDOW(preedit_window), x, y);
4942 }
4943 
4944     static void
4945 im_preedit_window_open()
4946 {
4947     char *preedit_string;
4948 #if !GTK_CHECK_VERSION(3,16,0)
4949     char buf[8];
4950 #endif
4951     PangoAttrList *attr_list;
4952     PangoLayout *layout;
4953 #if GTK_CHECK_VERSION(3,0,0)
4954 # if !GTK_CHECK_VERSION(3,16,0)
4955     GdkRGBA color;
4956 # endif
4957 #else
4958     GdkColor color;
4959 #endif
4960     gint w, h;
4961 
4962     if (preedit_window == NULL)
4963     {
4964 	preedit_window = gtk_window_new(GTK_WINDOW_POPUP);
4965 	gtk_window_set_transient_for(GTK_WINDOW(preedit_window),
4966 						     GTK_WINDOW(gui.mainwin));
4967 	preedit_label = gtk_label_new("");
4968 	gtk_widget_set_name(preedit_label, "vim-gui-preedit-area");
4969 	gtk_container_add(GTK_CONTAINER(preedit_window), preedit_label);
4970     }
4971 
4972 #if GTK_CHECK_VERSION(3,16,0)
4973     {
4974 	GtkStyleContext * const context
4975 				  = gtk_widget_get_style_context(gui.drawarea);
4976 	GtkCssProvider * const provider = gtk_css_provider_new();
4977 	gchar		   *css = NULL;
4978 	const char * const fontname
4979 			   = pango_font_description_get_family(gui.norm_font);
4980 	gint	fontsize
4981 		= pango_font_description_get_size(gui.norm_font) / PANGO_SCALE;
4982 	gchar	*fontsize_propval = NULL;
4983 
4984 	if (!pango_font_description_get_size_is_absolute(gui.norm_font))
4985 	{
4986 	    /* fontsize was given in points.  Convert it into that in pixels
4987 	     * to use with CSS. */
4988 	    GdkScreen * const screen
4989 		  = gdk_window_get_screen(gtk_widget_get_window(gui.mainwin));
4990 	    const gdouble dpi = gdk_screen_get_resolution(screen);
4991 	    fontsize = dpi * fontsize / 72;
4992 	}
4993 	if (fontsize > 0)
4994 	    fontsize_propval = g_strdup_printf("%dpx", fontsize);
4995 	else
4996 	    fontsize_propval = g_strdup_printf("inherit");
4997 
4998 	css = g_strdup_printf(
4999 		"widget#vim-gui-preedit-area {\n"
5000 		"  font-family: %s,monospace;\n"
5001 		"  font-size: %s;\n"
5002 		"  color: #%.2lx%.2lx%.2lx;\n"
5003 		"  background-color: #%.2lx%.2lx%.2lx;\n"
5004 		"}\n",
5005 		fontname != NULL ? fontname : "inherit",
5006 		fontsize_propval,
5007 		(gui.norm_pixel >> 16) & 0xff,
5008 		(gui.norm_pixel >> 8) & 0xff,
5009 		gui.norm_pixel & 0xff,
5010 		(gui.back_pixel >> 16) & 0xff,
5011 		(gui.back_pixel >> 8) & 0xff,
5012 		gui.back_pixel & 0xff);
5013 
5014 	gtk_css_provider_load_from_data(provider, css, -1, NULL);
5015 	gtk_style_context_add_provider(context,
5016 				     GTK_STYLE_PROVIDER(provider), G_MAXUINT);
5017 
5018 	g_free(css);
5019 	g_free(fontsize_propval);
5020 	g_object_unref(provider);
5021     }
5022 #elif GTK_CHECK_VERSION(3,0,0)
5023     gtk_widget_override_font(preedit_label, gui.norm_font);
5024 
5025     vim_snprintf(buf, sizeof(buf), "#%06X", gui.norm_pixel);
5026     gdk_rgba_parse(&color, buf);
5027     gtk_widget_override_color(preedit_label, GTK_STATE_FLAG_NORMAL, &color);
5028 
5029     vim_snprintf(buf, sizeof(buf), "#%06X", gui.back_pixel);
5030     gdk_rgba_parse(&color, buf);
5031     gtk_widget_override_background_color(preedit_label, GTK_STATE_FLAG_NORMAL,
5032 								      &color);
5033 #else
5034     gtk_widget_modify_font(preedit_label, gui.norm_font);
5035 
5036     vim_snprintf(buf, sizeof(buf), "#%06X", (unsigned)gui.norm_pixel);
5037     gdk_color_parse(buf, &color);
5038     gtk_widget_modify_fg(preedit_label, GTK_STATE_NORMAL, &color);
5039 
5040     vim_snprintf(buf, sizeof(buf), "#%06X", (unsigned)gui.back_pixel);
5041     gdk_color_parse(buf, &color);
5042     gtk_widget_modify_bg(preedit_window, GTK_STATE_NORMAL, &color);
5043 #endif
5044 
5045     gtk_im_context_get_preedit_string(xic, &preedit_string, &attr_list, NULL);
5046 
5047     if (preedit_string[0] != NUL)
5048     {
5049 	gtk_label_set_text(GTK_LABEL(preedit_label), preedit_string);
5050 	gtk_label_set_attributes(GTK_LABEL(preedit_label), attr_list);
5051 
5052 	layout = gtk_label_get_layout(GTK_LABEL(preedit_label));
5053 	pango_layout_get_pixel_size(layout, &w, &h);
5054 	h = MAX(h, gui.char_height);
5055 	gtk_window_resize(GTK_WINDOW(preedit_window), w, h);
5056 
5057 	gtk_widget_show_all(preedit_window);
5058 
5059 	im_preedit_window_set_position();
5060     }
5061 
5062     g_free(preedit_string);
5063     pango_attr_list_unref(attr_list);
5064 }
5065 
5066     static void
5067 im_preedit_window_close()
5068 {
5069     if (preedit_window != NULL)
5070 	gtk_widget_hide(preedit_window);
5071 }
5072 
5073     static void
5074 im_show_preedit()
5075 {
5076     im_preedit_window_open();
5077 
5078     if (p_mh) /* blank out the pointer if necessary */
5079 	gui_mch_mousehide(TRUE);
5080 }
5081 
5082     static void
5083 im_delete_preedit(void)
5084 {
5085     char_u bskey[]  = {CSI, 'k', 'b'};
5086     char_u delkey[] = {CSI, 'k', 'D'};
5087 
5088     if (p_imst == IM_OVER_THE_SPOT)
5089     {
5090 	im_preedit_window_close();
5091 	return;
5092     }
5093 
5094     if (State & NORMAL
5095 #ifdef FEAT_TERMINAL
5096 	    && !term_use_loop()
5097 #endif
5098        )
5099     {
5100 	im_preedit_cursor = 0;
5101 	return;
5102     }
5103     for (; im_preedit_cursor > 0; --im_preedit_cursor)
5104 	add_to_input_buf(bskey, (int)sizeof(bskey));
5105 
5106     for (; im_preedit_trailing > 0; --im_preedit_trailing)
5107 	add_to_input_buf(delkey, (int)sizeof(delkey));
5108 }
5109 
5110 /*
5111  * Move the cursor left by "num_move_back" characters.
5112  * Note that ins_left() checks im_is_preediting() to avoid breaking undo for
5113  * these K_LEFT keys.
5114  */
5115     static void
5116 im_correct_cursor(int num_move_back)
5117 {
5118     char_u backkey[] = {CSI, 'k', 'l'};
5119 
5120     if (State & NORMAL)
5121 	return;
5122 #  ifdef FEAT_RIGHTLEFT
5123     if ((State & CMDLINE) == 0 && curwin != NULL && curwin->w_p_rl)
5124 	backkey[2] = 'r';
5125 #  endif
5126     for (; num_move_back > 0; --num_move_back)
5127 	add_to_input_buf(backkey, (int)sizeof(backkey));
5128 }
5129 
5130 static int xim_expected_char = NUL;
5131 static int xim_ignored_char = FALSE;
5132 
5133 /*
5134  * Update the mode and cursor while in an IM callback.
5135  */
5136     static void
5137 im_show_info(void)
5138 {
5139     int	    old_vgetc_busy;
5140 
5141     old_vgetc_busy = vgetc_busy;
5142     vgetc_busy = TRUE;
5143     showmode();
5144     vgetc_busy = old_vgetc_busy;
5145     if ((State & NORMAL) || (State & INSERT))
5146 	setcursor();
5147     out_flush();
5148 }
5149 
5150 /*
5151  * Callback invoked when the user finished preediting.
5152  * Put the final string into the input buffer.
5153  */
5154     static void
5155 im_commit_cb(GtkIMContext *context UNUSED,
5156 	     const gchar *str,
5157 	     gpointer data UNUSED)
5158 {
5159     int		slen = (int)STRLEN(str);
5160     int		add_to_input = TRUE;
5161     int		clen;
5162     int		len = slen;
5163     int		commit_with_preedit = TRUE;
5164     char_u	*im_str;
5165 
5166 #ifdef XIM_DEBUG
5167     xim_log("im_commit_cb(): %s\n", str);
5168 #endif
5169 
5170     if (p_imst == IM_ON_THE_SPOT)
5171     {
5172 	/* The imhangul module doesn't reset the preedit string before
5173 	 * committing.  Call im_delete_preedit() to work around that. */
5174 	im_delete_preedit();
5175 
5176 	/* Indicate that preediting has finished. */
5177 	if (preedit_start_col == MAXCOL)
5178 	{
5179 	    init_preedit_start_col();
5180 	    commit_with_preedit = FALSE;
5181 	}
5182 
5183 	/* The thing which setting "preedit_start_col" to MAXCOL means that
5184 	 * "preedit_start_col" will be set forcedly when calling
5185 	 * preedit_changed_cb() next time.
5186 	 * "preedit_start_col" should not reset with MAXCOL on this part. Vim
5187 	 * is simulating the preediting by using add_to_input_str(). when
5188 	 * preedit begin immediately before committed, the typebuf is not
5189 	 * flushed to screen, then it can't get correct "preedit_start_col".
5190 	 * Thus, it should calculate the cells by adding cells of the committed
5191 	 * string. */
5192 	if (input_conv.vc_type != CONV_NONE)
5193 	{
5194 	    im_str = string_convert(&input_conv, (char_u *)str, &len);
5195 	    g_return_if_fail(im_str != NULL);
5196 	}
5197 	else
5198 	    im_str = (char_u *)str;
5199 
5200 	clen = mb_string2cells(im_str, len);
5201 
5202 	if (input_conv.vc_type != CONV_NONE)
5203 	    vim_free(im_str);
5204 	preedit_start_col += clen;
5205     }
5206 
5207     /* Is this a single character that matches a keypad key that's just
5208      * been pressed?  If so, we don't want it to be entered as such - let
5209      * us carry on processing the raw keycode so that it may be used in
5210      * mappings as <kSomething>. */
5211     if (xim_expected_char != NUL)
5212     {
5213 	/* We're currently processing a keypad or other special key */
5214 	if (slen == 1 && str[0] == xim_expected_char)
5215 	{
5216 	    /* It's a match - don't do it here */
5217 	    xim_ignored_char = TRUE;
5218 	    add_to_input = FALSE;
5219 	}
5220 	else
5221 	{
5222 	    /* Not a match */
5223 	    xim_ignored_char = FALSE;
5224 	}
5225     }
5226 
5227     if (add_to_input)
5228 	im_add_to_input((char_u *)str, slen);
5229 
5230     if (p_imst == IM_ON_THE_SPOT)
5231     {
5232 	/* Inserting chars while "im_is_active" is set does not cause a
5233 	 * change of buffer.  When the chars are committed the buffer must be
5234 	 * marked as changed. */
5235 	if (!commit_with_preedit)
5236 	    preedit_start_col = MAXCOL;
5237 
5238 	/* This flag is used in changed() at next call. */
5239 	xim_changed_while_preediting = TRUE;
5240     }
5241 
5242     if (gtk_main_level() > 0)
5243 	gtk_main_quit();
5244 }
5245 
5246 /*
5247  * Callback invoked after start to the preedit.
5248  */
5249     static void
5250 im_preedit_start_cb(GtkIMContext *context UNUSED, gpointer data UNUSED)
5251 {
5252 #ifdef XIM_DEBUG
5253     xim_log("im_preedit_start_cb()\n");
5254 #endif
5255 
5256     im_is_active = TRUE;
5257     preedit_is_active = TRUE;
5258     gui_update_cursor(TRUE, FALSE);
5259     im_show_info();
5260 }
5261 
5262 /*
5263  * Callback invoked after end to the preedit.
5264  */
5265     static void
5266 im_preedit_end_cb(GtkIMContext *context UNUSED, gpointer data UNUSED)
5267 {
5268 #ifdef XIM_DEBUG
5269     xim_log("im_preedit_end_cb()\n");
5270 #endif
5271     im_delete_preedit();
5272 
5273     /* Indicate that preediting has finished */
5274     if (p_imst == IM_ON_THE_SPOT)
5275 	preedit_start_col = MAXCOL;
5276     xim_has_preediting = FALSE;
5277 
5278 #if 0
5279     /* Removal of this line suggested by Takuhiro Nishioka.  Fixes that IM was
5280      * switched off unintentionally.  We now use preedit_is_active (added by
5281      * SungHyun Nam). */
5282     im_is_active = FALSE;
5283 #endif
5284     preedit_is_active = FALSE;
5285     gui_update_cursor(TRUE, FALSE);
5286     im_show_info();
5287 }
5288 
5289 /*
5290  * Callback invoked after changes to the preedit string.  If the preedit
5291  * string was empty before, remember the preedit start column so we know
5292  * where to apply feedback attributes.  Delete the previous preedit string
5293  * if there was one, save the new preedit cursor offset, and put the new
5294  * string into the input buffer.
5295  *
5296  * TODO: The pragmatic "put into input buffer" approach used here has
5297  *       several fundamental problems:
5298  *
5299  * - The characters in the preedit string are subject to remapping.
5300  *   That's broken, only the finally committed string should be remapped.
5301  *
5302  * - There is a race condition involved:  The retrieved value for the
5303  *   current cursor position will be wrong if any unprocessed characters
5304  *   are still queued in the input buffer.
5305  *
5306  * - Due to the lack of synchronization between the file buffer in memory
5307  *   and any typed characters, it's practically impossible to implement the
5308  *   "retrieve_surrounding" and "delete_surrounding" signals reliably.  IM
5309  *   modules for languages such as Thai are likely to rely on this feature
5310  *   for proper operation.
5311  *
5312  * Conclusions:  I think support for preediting needs to be moved to the
5313  * core parts of Vim.  Ideally, until it has been committed, the preediting
5314  * string should only be displayed and not affect the buffer content at all.
5315  * The question how to deal with the synchronization issue still remains.
5316  * Circumventing the input buffer is probably not desirable.  Anyway, I think
5317  * implementing "retrieve_surrounding" is the only hard problem.
5318  *
5319  * One way to solve all of this in a clean manner would be to queue all key
5320  * press/release events "as is" in the input buffer, and apply the IM filtering
5321  * at the receiving end of the queue.  This, however, would have a rather large
5322  * impact on the code base.  If there is an easy way to force processing of all
5323  * remaining input from within the "retrieve_surrounding" signal handler, this
5324  * might not be necessary.  Gotta ask on vim-dev for opinions.
5325  */
5326     static void
5327 im_preedit_changed_cb(GtkIMContext *context, gpointer data UNUSED)
5328 {
5329     char    *preedit_string = NULL;
5330     int	    cursor_index    = 0;
5331     int	    num_move_back   = 0;
5332     char_u  *str;
5333     char_u  *p;
5334     int	    i;
5335 
5336     if (p_imst == IM_ON_THE_SPOT)
5337 	gtk_im_context_get_preedit_string(context,
5338 					  &preedit_string, NULL,
5339 					  &cursor_index);
5340     else
5341 	gtk_im_context_get_preedit_string(context,
5342 					  &preedit_string, NULL,
5343 					  NULL);
5344 
5345 #ifdef XIM_DEBUG
5346     xim_log("im_preedit_changed_cb(): %s\n", preedit_string);
5347 #endif
5348 
5349     g_return_if_fail(preedit_string != NULL); /* just in case */
5350 
5351     if (p_imst == IM_OVER_THE_SPOT)
5352     {
5353 	if (preedit_string[0] == NUL)
5354 	{
5355 	    xim_has_preediting = FALSE;
5356 	    im_delete_preedit();
5357 	}
5358 	else
5359 	{
5360 	    xim_has_preediting = TRUE;
5361 	    im_show_preedit();
5362 	}
5363     }
5364     else
5365     {
5366 	/* If preedit_start_col is MAXCOL set it to the current cursor position. */
5367 	if (preedit_start_col == MAXCOL && preedit_string[0] != '\0')
5368 	{
5369 	    xim_has_preediting = TRUE;
5370 
5371 	    /* Urgh, this breaks if the input buffer isn't empty now */
5372 	    init_preedit_start_col();
5373 	}
5374 	else if (cursor_index == 0 && preedit_string[0] == '\0')
5375 	{
5376 	    xim_has_preediting = FALSE;
5377 
5378 	    /* If at the start position (after typing backspace)
5379 	     * preedit_start_col must be reset. */
5380 	    preedit_start_col = MAXCOL;
5381 	}
5382 
5383 	im_delete_preedit();
5384 
5385 	/*
5386 	 * Compute the end of the preediting area: "preedit_end_col".
5387 	 * According to the documentation of gtk_im_context_get_preedit_string(),
5388 	 * the cursor_pos output argument returns the offset in bytes.  This is
5389 	 * unfortunately not true -- real life shows the offset is in characters,
5390 	 * and the GTK+ source code agrees with me.  Will file a bug later.
5391 	 */
5392 	if (preedit_start_col != MAXCOL)
5393 	    preedit_end_col = preedit_start_col;
5394 	str = (char_u *)preedit_string;
5395 	for (p = str, i = 0; *p != NUL; p += utf_byte2len(*p), ++i)
5396 	{
5397 	    int is_composing;
5398 
5399 	    is_composing = ((*p & 0x80) != 0 && utf_iscomposing(utf_ptr2char(p)));
5400 	    /*
5401 	     * These offsets are used as counters when generating <BS> and <Del>
5402 	     * to delete the preedit string.  So don't count composing characters
5403 	     * unless 'delcombine' is enabled.
5404 	     */
5405 	    if (!is_composing || p_deco)
5406 	    {
5407 		if (i < cursor_index)
5408 		    ++im_preedit_cursor;
5409 		else
5410 		    ++im_preedit_trailing;
5411 	    }
5412 	    if (!is_composing && i >= cursor_index)
5413 	    {
5414 		/* This is essentially the same as im_preedit_trailing, except
5415 		 * composing characters are not counted even if p_deco is set. */
5416 		++num_move_back;
5417 	    }
5418 	    if (preedit_start_col != MAXCOL)
5419 		preedit_end_col += utf_ptr2cells(p);
5420 	}
5421 
5422 	if (p > str)
5423 	{
5424 	    im_add_to_input(str, (int)(p - str));
5425 	    im_correct_cursor(num_move_back);
5426 	}
5427     }
5428 
5429     g_free(preedit_string);
5430 
5431     if (gtk_main_level() > 0)
5432 	gtk_main_quit();
5433 }
5434 
5435 /*
5436  * Translate the Pango attributes at iter to Vim highlighting attributes.
5437  * Ignore attributes not supported by Vim highlighting.  This shouldn't have
5438  * too much impact -- right now we handle even more attributes than necessary
5439  * for the IM modules I tested with.
5440  */
5441     static int
5442 translate_pango_attributes(PangoAttrIterator *iter)
5443 {
5444     PangoAttribute  *attr;
5445     int		    char_attr = HL_NORMAL;
5446 
5447     attr = pango_attr_iterator_get(iter, PANGO_ATTR_UNDERLINE);
5448     if (attr != NULL && ((PangoAttrInt *)attr)->value
5449 						 != (int)PANGO_UNDERLINE_NONE)
5450 	char_attr |= HL_UNDERLINE;
5451 
5452     attr = pango_attr_iterator_get(iter, PANGO_ATTR_WEIGHT);
5453     if (attr != NULL && ((PangoAttrInt *)attr)->value >= (int)PANGO_WEIGHT_BOLD)
5454 	char_attr |= HL_BOLD;
5455 
5456     attr = pango_attr_iterator_get(iter, PANGO_ATTR_STYLE);
5457     if (attr != NULL && ((PangoAttrInt *)attr)->value
5458 						   != (int)PANGO_STYLE_NORMAL)
5459 	char_attr |= HL_ITALIC;
5460 
5461     attr = pango_attr_iterator_get(iter, PANGO_ATTR_BACKGROUND);
5462     if (attr != NULL)
5463     {
5464 	const PangoColor *color = &((PangoAttrColor *)attr)->color;
5465 
5466 	/* Assume inverse if black background is requested */
5467 	if ((color->red | color->green | color->blue) == 0)
5468 	    char_attr |= HL_INVERSE;
5469     }
5470 
5471     return char_attr;
5472 }
5473 
5474 /*
5475  * Retrieve the highlighting attributes at column col in the preedit string.
5476  * Return -1 if not in preediting mode or if col is out of range.
5477  */
5478     int
5479 im_get_feedback_attr(int col)
5480 {
5481     char	    *preedit_string = NULL;
5482     PangoAttrList   *attr_list	    = NULL;
5483     int		    char_attr	    = -1;
5484 
5485     if (xic == NULL)
5486 	return char_attr;
5487 
5488     gtk_im_context_get_preedit_string(xic, &preedit_string, &attr_list, NULL);
5489 
5490     if (preedit_string != NULL && attr_list != NULL)
5491     {
5492 	int idx;
5493 
5494 	/* Get the byte index as used by PangoAttrIterator */
5495 	for (idx = 0; col > 0 && preedit_string[idx] != '\0'; --col)
5496 	    idx += utfc_ptr2len((char_u *)preedit_string + idx);
5497 
5498 	if (preedit_string[idx] != '\0')
5499 	{
5500 	    PangoAttrIterator	*iter;
5501 	    int			start, end;
5502 
5503 	    char_attr = HL_NORMAL;
5504 	    iter = pango_attr_list_get_iterator(attr_list);
5505 
5506 	    /* Extract all relevant attributes from the list. */
5507 	    do
5508 	    {
5509 		pango_attr_iterator_range(iter, &start, &end);
5510 
5511 		if (idx >= start && idx < end)
5512 		    char_attr |= translate_pango_attributes(iter);
5513 	    }
5514 	    while (pango_attr_iterator_next(iter));
5515 
5516 	    pango_attr_iterator_destroy(iter);
5517 	}
5518     }
5519 
5520     if (attr_list != NULL)
5521 	pango_attr_list_unref(attr_list);
5522     g_free(preedit_string);
5523 
5524     return char_attr;
5525 }
5526 
5527     void
5528 xim_init(void)
5529 {
5530 #ifdef XIM_DEBUG
5531     xim_log("xim_init()\n");
5532 #endif
5533 
5534     g_return_if_fail(gui.drawarea != NULL);
5535     g_return_if_fail(gtk_widget_get_window(gui.drawarea) != NULL);
5536 
5537     xic = gtk_im_multicontext_new();
5538     g_object_ref(xic);
5539 
5540     im_commit_handler_id = g_signal_connect(G_OBJECT(xic), "commit",
5541 					    G_CALLBACK(&im_commit_cb), NULL);
5542     g_signal_connect(G_OBJECT(xic), "preedit_changed",
5543 		     G_CALLBACK(&im_preedit_changed_cb), NULL);
5544     g_signal_connect(G_OBJECT(xic), "preedit_start",
5545 		     G_CALLBACK(&im_preedit_start_cb), NULL);
5546     g_signal_connect(G_OBJECT(xic), "preedit_end",
5547 		     G_CALLBACK(&im_preedit_end_cb), NULL);
5548 
5549     gtk_im_context_set_client_window(xic, gtk_widget_get_window(gui.drawarea));
5550 }
5551 
5552     void
5553 im_shutdown(void)
5554 {
5555 #ifdef XIM_DEBUG
5556     xim_log("im_shutdown()\n");
5557 #endif
5558 
5559     if (xic != NULL)
5560     {
5561 	gtk_im_context_focus_out(xic);
5562 	g_object_unref(xic);
5563 	xic = NULL;
5564     }
5565     im_is_active = FALSE;
5566     im_commit_handler_id = 0;
5567     if (p_imst == IM_ON_THE_SPOT)
5568 	preedit_start_col = MAXCOL;
5569     xim_has_preediting = FALSE;
5570 }
5571 
5572 /*
5573  * Convert the string argument to keyval and state for GdkEventKey.
5574  * If str is valid return TRUE, otherwise FALSE.
5575  *
5576  * See 'imactivatekey' for documentation of the format.
5577  */
5578     static int
5579 im_string_to_keyval(const char *str, unsigned int *keyval, unsigned int *state)
5580 {
5581     const char	    *mods_end;
5582     unsigned	    tmp_keyval;
5583     unsigned	    tmp_state = 0;
5584 
5585     mods_end = strrchr(str, '-');
5586     mods_end = (mods_end != NULL) ? mods_end + 1 : str;
5587 
5588     /* Parse modifier keys */
5589     while (str < mods_end)
5590 	switch (*str++)
5591 	{
5592 	    case '-':							break;
5593 	    case 'S': case 's': tmp_state |= (unsigned)GDK_SHIFT_MASK;	break;
5594 	    case 'L': case 'l': tmp_state |= (unsigned)GDK_LOCK_MASK;	break;
5595 	    case 'C': case 'c': tmp_state |= (unsigned)GDK_CONTROL_MASK;break;
5596 	    case '1':		tmp_state |= (unsigned)GDK_MOD1_MASK;	break;
5597 	    case '2':		tmp_state |= (unsigned)GDK_MOD2_MASK;	break;
5598 	    case '3':		tmp_state |= (unsigned)GDK_MOD3_MASK;	break;
5599 	    case '4':		tmp_state |= (unsigned)GDK_MOD4_MASK;	break;
5600 	    case '5':		tmp_state |= (unsigned)GDK_MOD5_MASK;	break;
5601 	    default:
5602 		return FALSE;
5603 	}
5604 
5605     tmp_keyval = gdk_keyval_from_name(str);
5606 
5607     if (tmp_keyval == 0 || tmp_keyval == GDK_VoidSymbol)
5608 	return FALSE;
5609 
5610     if (keyval != NULL)
5611 	*keyval = tmp_keyval;
5612     if (state != NULL)
5613 	*state = tmp_state;
5614 
5615     return TRUE;
5616 }
5617 
5618 /*
5619  * Return TRUE if p_imak is valid, otherwise FALSE.  As a special case, an
5620  * empty string is also regarded as valid.
5621  *
5622  * Note: The numerical key value of p_imak is cached if it was valid; thus
5623  * boldly assuming im_xim_isvalid_imactivate() will always be called whenever
5624  * 'imak' changes.  This is currently the case but not obvious -- should
5625  * probably rename the function for clarity.
5626  */
5627     int
5628 im_xim_isvalid_imactivate(void)
5629 {
5630     if (p_imak[0] == NUL)
5631     {
5632 	im_activatekey_keyval = GDK_VoidSymbol;
5633 	im_activatekey_state  = 0;
5634 	return TRUE;
5635     }
5636 
5637     return im_string_to_keyval((const char *)p_imak,
5638 			       &im_activatekey_keyval,
5639 			       &im_activatekey_state);
5640 }
5641 
5642     static void
5643 im_synthesize_keypress(unsigned int keyval, unsigned int state)
5644 {
5645     GdkEventKey *event;
5646 
5647     event = (GdkEventKey *)gdk_event_new(GDK_KEY_PRESS);
5648     g_object_ref(gtk_widget_get_window(gui.drawarea));
5649 					/* unreffed by gdk_event_free() */
5650     event->window = gtk_widget_get_window(gui.drawarea);
5651     event->send_event = TRUE;
5652     event->time = GDK_CURRENT_TIME;
5653     event->state  = state;
5654     event->keyval = keyval;
5655     event->hardware_keycode = /* needed for XIM */
5656 	XKeysymToKeycode(GDK_WINDOW_XDISPLAY(event->window), (KeySym)keyval);
5657     event->length = 0;
5658     event->string = NULL;
5659 
5660     gtk_im_context_filter_keypress(xic, event);
5661 
5662     /* For consistency, also send the corresponding release event. */
5663     event->type = GDK_KEY_RELEASE;
5664     event->send_event = FALSE;
5665     gtk_im_context_filter_keypress(xic, event);
5666 
5667     gdk_event_free((GdkEvent *)event);
5668 }
5669 
5670     void
5671 xim_reset(void)
5672 {
5673 # ifdef FEAT_EVAL
5674     if (USE_IMACTIVATEFUNC)
5675 	call_imactivatefunc(im_is_active);
5676     else
5677 # endif
5678     if (xic != NULL)
5679     {
5680 	gtk_im_context_reset(xic);
5681 
5682 	if (p_imdisable)
5683 	    im_shutdown();
5684 	else
5685 	{
5686 	    xim_set_focus(gui.in_focus);
5687 
5688 	    if (im_activatekey_keyval != GDK_VoidSymbol)
5689 	    {
5690 		if (im_is_active)
5691 		{
5692 		    g_signal_handler_block(xic, im_commit_handler_id);
5693 		    im_synthesize_keypress(im_activatekey_keyval,
5694 						    im_activatekey_state);
5695 		    g_signal_handler_unblock(xic, im_commit_handler_id);
5696 		}
5697 	    }
5698 	    else
5699 	    {
5700 		im_shutdown();
5701 		xim_init();
5702 		xim_set_focus(gui.in_focus);
5703 	    }
5704 	}
5705     }
5706 
5707     if (p_imst == IM_ON_THE_SPOT)
5708 	preedit_start_col = MAXCOL;
5709     xim_has_preediting = FALSE;
5710 }
5711 
5712     int
5713 xim_queue_key_press_event(GdkEventKey *event, int down)
5714 {
5715     if (down)
5716     {
5717 	/*
5718 	 * Workaround GTK2 XIM 'feature' that always converts keypad keys to
5719 	 * chars., even when not part of an IM sequence (ref. feature of
5720 	 * gdk/gdkkeyuni.c).
5721 	 * Flag any keypad keys that might represent a single char.
5722 	 * If this (on its own - i.e., not part of an IM sequence) is
5723 	 * committed while we're processing one of these keys, we can ignore
5724 	 * that commit and go ahead & process it ourselves.  That way we can
5725 	 * still distinguish keypad keys for use in mappings.
5726 	 * Also add GDK_space to make <S-Space> work.
5727 	 */
5728 	switch (event->keyval)
5729 	{
5730 	    case GDK_KP_Add:      xim_expected_char = '+';  break;
5731 	    case GDK_KP_Subtract: xim_expected_char = '-';  break;
5732 	    case GDK_KP_Divide:   xim_expected_char = '/';  break;
5733 	    case GDK_KP_Multiply: xim_expected_char = '*';  break;
5734 	    case GDK_KP_Decimal:  xim_expected_char = '.';  break;
5735 	    case GDK_KP_Equal:    xim_expected_char = '=';  break;
5736 	    case GDK_KP_0:	  xim_expected_char = '0';  break;
5737 	    case GDK_KP_1:	  xim_expected_char = '1';  break;
5738 	    case GDK_KP_2:	  xim_expected_char = '2';  break;
5739 	    case GDK_KP_3:	  xim_expected_char = '3';  break;
5740 	    case GDK_KP_4:	  xim_expected_char = '4';  break;
5741 	    case GDK_KP_5:	  xim_expected_char = '5';  break;
5742 	    case GDK_KP_6:	  xim_expected_char = '6';  break;
5743 	    case GDK_KP_7:	  xim_expected_char = '7';  break;
5744 	    case GDK_KP_8:	  xim_expected_char = '8';  break;
5745 	    case GDK_KP_9:	  xim_expected_char = '9';  break;
5746 	    case GDK_space:	  xim_expected_char = ' ';  break;
5747 	    default:		  xim_expected_char = NUL;
5748 	}
5749 	xim_ignored_char = FALSE;
5750     }
5751 
5752     /*
5753      * When typing fFtT, XIM may be activated. Thus it must pass
5754      * gtk_im_context_filter_keypress() in Normal mode.
5755      * And while doing :sh too.
5756      */
5757     if (xic != NULL && !p_imdisable
5758 		    && (State & (INSERT | CMDLINE | NORMAL | EXTERNCMD)) != 0)
5759     {
5760 	/*
5761 	 * Filter 'imactivatekey' and map it to CTRL-^.  This way, Vim is
5762 	 * always aware of the current status of IM, and can even emulate
5763 	 * the activation key for modules that don't support one.
5764 	 */
5765 	if (event->keyval == im_activatekey_keyval
5766 	     && (event->state & im_activatekey_state) == im_activatekey_state)
5767 	{
5768 	    unsigned int state_mask;
5769 
5770 	    /* Require the state of the 3 most used modifiers to match exactly.
5771 	     * Otherwise e.g. <S-C-space> would be unusable for other purposes
5772 	     * if the IM activate key is <S-space>. */
5773 	    state_mask  = im_activatekey_state;
5774 	    state_mask |= ((int)GDK_SHIFT_MASK | (int)GDK_CONTROL_MASK
5775 							| (int)GDK_MOD1_MASK);
5776 
5777 	    if ((event->state & state_mask) != im_activatekey_state)
5778 		return FALSE;
5779 
5780 	    /* Don't send it a second time on GDK_KEY_RELEASE. */
5781 	    if (event->type != GDK_KEY_PRESS)
5782 		return TRUE;
5783 
5784 	    if (map_to_exists_mode((char_u *)"", LANGMAP, FALSE))
5785 	    {
5786 		im_set_active(FALSE);
5787 
5788 		/* ":lmap" mappings exists, toggle use of mappings. */
5789 		State ^= LANGMAP;
5790 		if (State & LANGMAP)
5791 		{
5792 		    curbuf->b_p_iminsert = B_IMODE_NONE;
5793 		    State &= ~LANGMAP;
5794 		}
5795 		else
5796 		{
5797 		    curbuf->b_p_iminsert = B_IMODE_LMAP;
5798 		    State |= LANGMAP;
5799 		}
5800 		return TRUE;
5801 	    }
5802 
5803 	    return gtk_im_context_filter_keypress(xic, event);
5804 	}
5805 
5806 	/* Don't filter events through the IM context if IM isn't active
5807 	 * right now.  Unlike with GTK+ 1.2 we cannot rely on the IM module
5808 	 * not doing anything before the activation key was sent. */
5809 	if (im_activatekey_keyval == GDK_VoidSymbol || im_is_active)
5810 	{
5811 	    int imresult = gtk_im_context_filter_keypress(xic, event);
5812 
5813 	    if (p_imst == IM_ON_THE_SPOT)
5814 	    {
5815 		/* Some XIM send following sequence:
5816 		 * 1. preedited string.
5817 		 * 2. committed string.
5818 		 * 3. line changed key.
5819 		 * 4. preedited string.
5820 		 * 5. remove preedited string.
5821 		 * if 3, Vim can't move back the above line for 5.
5822 		 * thus, this part should not parse the key. */
5823 		if (!imresult && preedit_start_col != MAXCOL
5824 					    && event->keyval == GDK_Return)
5825 		{
5826 		    im_synthesize_keypress(GDK_Return, 0U);
5827 		    return FALSE;
5828 		}
5829 	    }
5830 
5831 	    /* If XIM tried to commit a keypad key as a single char.,
5832 	     * ignore it so we can use the keypad key 'raw', for mappings. */
5833 	    if (xim_expected_char != NUL && xim_ignored_char)
5834 		/* We had a keypad key, and XIM tried to thieve it */
5835 		return FALSE;
5836 
5837 	    /* This is supposed to fix a problem with iBus, that space
5838 	     * characters don't work in input mode. */
5839 	    xim_expected_char = NUL;
5840 
5841 	    /* Normal processing */
5842 	    return imresult;
5843 	}
5844     }
5845 
5846     return FALSE;
5847 }
5848 
5849     int
5850 im_get_status(void)
5851 {
5852 #  ifdef FEAT_EVAL
5853     if (USE_IMSTATUSFUNC)
5854 	return call_imstatusfunc();
5855 #  endif
5856     return im_is_active;
5857 }
5858 
5859     int
5860 preedit_get_status(void)
5861 {
5862     return preedit_is_active;
5863 }
5864 
5865     int
5866 im_is_preediting(void)
5867 {
5868     return xim_has_preediting;
5869 }
5870 
5871 # else /* !FEAT_GUI_GTK */
5872 
5873 static int	xim_is_active = FALSE;  /* XIM should be active in the current
5874 					   mode */
5875 static int	xim_has_focus = FALSE;	/* XIM is really being used for Vim */
5876 #  ifdef FEAT_GUI_X11
5877 static XIMStyle	input_style;
5878 static int	status_area_enabled = TRUE;
5879 #  endif
5880 
5881 /*
5882  * Switch using XIM on/off.  This is used by the code that changes "State".
5883  * When 'imactivatefunc' is defined use that function instead.
5884  */
5885     void
5886 im_set_active(int active_arg)
5887 {
5888     int active = active_arg;
5889 
5890     /* If 'imdisable' is set, XIM is never active. */
5891     if (p_imdisable)
5892 	active = FALSE;
5893     else if (input_style & XIMPreeditPosition)
5894 	/* There is a problem in switching XIM off when preediting is used,
5895 	 * and it is not clear how this can be solved.  For now, keep XIM on
5896 	 * all the time, like it was done in Vim 5.8. */
5897 	active = TRUE;
5898 
5899 #  if defined(FEAT_EVAL)
5900     if (USE_IMACTIVATEFUNC)
5901     {
5902 	if (active != im_get_status())
5903 	{
5904 	    call_imactivatefunc(active);
5905 	    xim_has_focus = active;
5906 	}
5907 	return;
5908     }
5909 #  endif
5910 
5911     if (xic == NULL)
5912 	return;
5913 
5914     /* Remember the active state, it is needed when Vim gets keyboard focus. */
5915     xim_is_active = active;
5916     xim_set_preedit();
5917 }
5918 
5919 /*
5920  * Adjust using XIM for gaining or losing keyboard focus.  Also called when
5921  * "xim_is_active" changes.
5922  */
5923     void
5924 xim_set_focus(int focus)
5925 {
5926     if (xic == NULL)
5927 	return;
5928 
5929     /*
5930      * XIM only gets focus when the Vim window has keyboard focus and XIM has
5931      * been set active for the current mode.
5932      */
5933     if (focus && xim_is_active)
5934     {
5935 	if (!xim_has_focus)
5936 	{
5937 	    xim_has_focus = TRUE;
5938 	    XSetICFocus(xic);
5939 	}
5940     }
5941     else
5942     {
5943 	if (xim_has_focus)
5944 	{
5945 	    xim_has_focus = FALSE;
5946 	    XUnsetICFocus(xic);
5947 	}
5948     }
5949 }
5950 
5951     void
5952 im_set_position(int row UNUSED, int col UNUSED)
5953 {
5954     xim_set_preedit();
5955 }
5956 
5957 /*
5958  * Set the XIM to the current cursor position.
5959  */
5960     void
5961 xim_set_preedit(void)
5962 {
5963     XVaNestedList attr_list;
5964     XRectangle spot_area;
5965     XPoint over_spot;
5966     int line_space;
5967 
5968     if (xic == NULL)
5969 	return;
5970 
5971     xim_set_focus(TRUE);
5972 
5973     if (!xim_has_focus)
5974     {
5975 	/* hide XIM cursor */
5976 	over_spot.x = 0;
5977 	over_spot.y = -100; /* arbitrary invisible position */
5978 	attr_list = (XVaNestedList) XVaCreateNestedList(0,
5979 							XNSpotLocation,
5980 							&over_spot,
5981 							NULL);
5982 	XSetICValues(xic, XNPreeditAttributes, attr_list, NULL);
5983 	XFree(attr_list);
5984 	return;
5985     }
5986 
5987     if (input_style & XIMPreeditPosition)
5988     {
5989 	if (xim_fg_color == INVALCOLOR)
5990 	{
5991 	    xim_fg_color = gui.def_norm_pixel;
5992 	    xim_bg_color = gui.def_back_pixel;
5993 	}
5994 	over_spot.x = TEXT_X(gui.col);
5995 	over_spot.y = TEXT_Y(gui.row);
5996 	spot_area.x = 0;
5997 	spot_area.y = 0;
5998 	spot_area.height = gui.char_height * Rows;
5999 	spot_area.width  = gui.char_width * Columns;
6000 	line_space = gui.char_height;
6001 	attr_list = (XVaNestedList) XVaCreateNestedList(0,
6002 					XNSpotLocation, &over_spot,
6003 					XNForeground, (Pixel) xim_fg_color,
6004 					XNBackground, (Pixel) xim_bg_color,
6005 					XNArea, &spot_area,
6006 					XNLineSpace, line_space,
6007 					NULL);
6008 	if (XSetICValues(xic, XNPreeditAttributes, attr_list, NULL))
6009 	    emsg(_("E284: Cannot set IC values"));
6010 	XFree(attr_list);
6011     }
6012 }
6013 
6014 #  if defined(FEAT_GUI_X11)
6015 static char e_xim[] = N_("E285: Failed to create input context");
6016 #  endif
6017 
6018 #  if defined(FEAT_GUI_X11) || defined(PROTO)
6019 #   if defined(XtSpecificationRelease) && XtSpecificationRelease >= 6 && !defined(SUN_SYSTEM)
6020 #    define USE_X11R6_XIM
6021 #   endif
6022 
6023 static int xim_real_init(Window x11_window, Display *x11_display);
6024 
6025 
6026 #  ifdef USE_X11R6_XIM
6027     static void
6028 xim_instantiate_cb(
6029     Display	*display,
6030     XPointer	client_data UNUSED,
6031     XPointer	call_data UNUSED)
6032 {
6033     Window	x11_window;
6034     Display	*x11_display;
6035 
6036 #   ifdef XIM_DEBUG
6037     xim_log("xim_instantiate_cb()\n");
6038 #   endif
6039 
6040     gui_get_x11_windis(&x11_window, &x11_display);
6041     if (display != x11_display)
6042 	return;
6043 
6044     xim_real_init(x11_window, x11_display);
6045     gui_set_shellsize(FALSE, FALSE, RESIZE_BOTH);
6046     if (xic != NULL)
6047 	XUnregisterIMInstantiateCallback(x11_display, NULL, NULL, NULL,
6048 					 xim_instantiate_cb, NULL);
6049 }
6050 
6051     static void
6052 xim_destroy_cb(
6053     XIM		im UNUSED,
6054     XPointer	client_data UNUSED,
6055     XPointer	call_data UNUSED)
6056 {
6057     Window	x11_window;
6058     Display	*x11_display;
6059 
6060 #   ifdef XIM_DEBUG
6061     xim_log("xim_destroy_cb()\n");
6062    #endif
6063     gui_get_x11_windis(&x11_window, &x11_display);
6064 
6065     xic = NULL;
6066     status_area_enabled = FALSE;
6067 
6068     gui_set_shellsize(FALSE, FALSE, RESIZE_BOTH);
6069 
6070     XRegisterIMInstantiateCallback(x11_display, NULL, NULL, NULL,
6071 				   xim_instantiate_cb, NULL);
6072 }
6073 #  endif
6074 
6075     void
6076 xim_init(void)
6077 {
6078     Window	x11_window;
6079     Display	*x11_display;
6080 
6081 #  ifdef XIM_DEBUG
6082     xim_log("xim_init()\n");
6083 #  endif
6084 
6085     gui_get_x11_windis(&x11_window, &x11_display);
6086 
6087     xic = NULL;
6088 
6089     if (xim_real_init(x11_window, x11_display))
6090 	return;
6091 
6092     gui_set_shellsize(FALSE, FALSE, RESIZE_BOTH);
6093 
6094 #  ifdef USE_X11R6_XIM
6095     XRegisterIMInstantiateCallback(x11_display, NULL, NULL, NULL,
6096 				   xim_instantiate_cb, NULL);
6097 #  endif
6098 }
6099 
6100     static int
6101 xim_real_init(Window x11_window, Display *x11_display)
6102 {
6103     int		i;
6104     char	*p,
6105 		*s,
6106 		*ns,
6107 		*end,
6108 		tmp[1024];
6109 #  define IMLEN_MAX 40
6110     char	buf[IMLEN_MAX + 7];
6111     XIM		xim = NULL;
6112     XIMStyles	*xim_styles;
6113     XIMStyle	this_input_style = 0;
6114     Boolean	found;
6115     XPoint	over_spot;
6116     XVaNestedList preedit_list, status_list;
6117 
6118     input_style = 0;
6119     status_area_enabled = FALSE;
6120 
6121     if (xic != NULL)
6122 	return FALSE;
6123 
6124     if (gui.rsrc_input_method != NULL && *gui.rsrc_input_method != NUL)
6125     {
6126 	strcpy(tmp, gui.rsrc_input_method);
6127 	for (ns = s = tmp; ns != NULL && *s != NUL;)
6128 	{
6129 	    s = (char *)skipwhite((char_u *)s);
6130 	    if (*s == NUL)
6131 		break;
6132 	    if ((ns = end = strchr(s, ',')) == NULL)
6133 		end = s + strlen(s);
6134 	    while (isspace(((char_u *)end)[-1]))
6135 		end--;
6136 	    *end = NUL;
6137 
6138 	    if (strlen(s) <= IMLEN_MAX)
6139 	    {
6140 		strcpy(buf, "@im=");
6141 		strcat(buf, s);
6142 		if ((p = XSetLocaleModifiers(buf)) != NULL && *p != NUL
6143 			&& (xim = XOpenIM(x11_display, NULL, NULL, NULL))
6144 								      != NULL)
6145 		    break;
6146 	    }
6147 
6148 	    s = ns + 1;
6149 	}
6150     }
6151 
6152     if (xim == NULL && (p = XSetLocaleModifiers("")) != NULL && *p != NUL)
6153 	xim = XOpenIM(x11_display, NULL, NULL, NULL);
6154 
6155     /* This is supposed to be useful to obtain characters through
6156      * XmbLookupString() without really using a XIM. */
6157     if (xim == NULL && (p = XSetLocaleModifiers("@im=none")) != NULL
6158 								 && *p != NUL)
6159 	xim = XOpenIM(x11_display, NULL, NULL, NULL);
6160 
6161     if (xim == NULL)
6162     {
6163 	/* Only give this message when verbose is set, because too many people
6164 	 * got this message when they didn't want to use a XIM. */
6165 	if (p_verbose > 0)
6166 	{
6167 	    verbose_enter();
6168 	    emsg(_("E286: Failed to open input method"));
6169 	    verbose_leave();
6170 	}
6171 	return FALSE;
6172     }
6173 
6174 #  ifdef USE_X11R6_XIM
6175     {
6176 	XIMCallback destroy_cb;
6177 
6178 	destroy_cb.callback = xim_destroy_cb;
6179 	destroy_cb.client_data = NULL;
6180 	if (XSetIMValues(xim, XNDestroyCallback, &destroy_cb, NULL))
6181 	    emsg(_("E287: Warning: Could not set destroy callback to IM"));
6182     }
6183 #  endif
6184 
6185     if (XGetIMValues(xim, XNQueryInputStyle, &xim_styles, NULL) || !xim_styles)
6186     {
6187 	emsg(_("E288: input method doesn't support any style"));
6188 	XCloseIM(xim);
6189 	return FALSE;
6190     }
6191 
6192     found = False;
6193     strcpy(tmp, gui.rsrc_preedit_type_name);
6194     for (s = tmp; s && !found; )
6195     {
6196 	while (*s && isspace((unsigned char)*s))
6197 	    s++;
6198 	if (!*s)
6199 	    break;
6200 	if ((ns = end = strchr(s, ',')) != 0)
6201 	    ns++;
6202 	else
6203 	    end = s + strlen(s);
6204 	while (isspace((unsigned char)*end))
6205 	    end--;
6206 	*end = '\0';
6207 
6208 	if (!strcmp(s, "OverTheSpot"))
6209 	    this_input_style = (XIMPreeditPosition | XIMStatusArea);
6210 	else if (!strcmp(s, "OffTheSpot"))
6211 	    this_input_style = (XIMPreeditArea | XIMStatusArea);
6212 	else if (!strcmp(s, "Root"))
6213 	    this_input_style = (XIMPreeditNothing | XIMStatusNothing);
6214 
6215 	for (i = 0; (unsigned short)i < xim_styles->count_styles; i++)
6216 	{
6217 	    if (this_input_style == xim_styles->supported_styles[i])
6218 	    {
6219 		found = True;
6220 		break;
6221 	    }
6222 	}
6223 	if (!found)
6224 	    for (i = 0; (unsigned short)i < xim_styles->count_styles; i++)
6225 	    {
6226 		if ((xim_styles->supported_styles[i] & this_input_style)
6227 			== (this_input_style & ~XIMStatusArea))
6228 		{
6229 		    this_input_style &= ~XIMStatusArea;
6230 		    found = True;
6231 		    break;
6232 		}
6233 	    }
6234 
6235 	s = ns;
6236     }
6237     XFree(xim_styles);
6238 
6239     if (!found)
6240     {
6241 	/* Only give this message when verbose is set, because too many people
6242 	 * got this message when they didn't want to use a XIM. */
6243 	if (p_verbose > 0)
6244 	{
6245 	    verbose_enter();
6246 	    emsg(_("E289: input method doesn't support my preedit type"));
6247 	    verbose_leave();
6248 	}
6249 	XCloseIM(xim);
6250 	return FALSE;
6251     }
6252 
6253     over_spot.x = TEXT_X(gui.col);
6254     over_spot.y = TEXT_Y(gui.row);
6255     input_style = this_input_style;
6256 
6257     /* A crash was reported when trying to pass gui.norm_font as XNFontSet,
6258      * thus that has been removed.  Hopefully the default works... */
6259 #  ifdef FEAT_XFONTSET
6260     if (gui.fontset != NOFONTSET)
6261     {
6262 	preedit_list = XVaCreateNestedList(0,
6263 				XNSpotLocation, &over_spot,
6264 				XNForeground, (Pixel)gui.def_norm_pixel,
6265 				XNBackground, (Pixel)gui.def_back_pixel,
6266 				XNFontSet, (XFontSet)gui.fontset,
6267 				NULL);
6268 	status_list = XVaCreateNestedList(0,
6269 				XNForeground, (Pixel)gui.def_norm_pixel,
6270 				XNBackground, (Pixel)gui.def_back_pixel,
6271 				XNFontSet, (XFontSet)gui.fontset,
6272 				NULL);
6273     }
6274     else
6275 #  endif
6276     {
6277 	preedit_list = XVaCreateNestedList(0,
6278 				XNSpotLocation, &over_spot,
6279 				XNForeground, (Pixel)gui.def_norm_pixel,
6280 				XNBackground, (Pixel)gui.def_back_pixel,
6281 				NULL);
6282 	status_list = XVaCreateNestedList(0,
6283 				XNForeground, (Pixel)gui.def_norm_pixel,
6284 				XNBackground, (Pixel)gui.def_back_pixel,
6285 				NULL);
6286     }
6287 
6288     xic = XCreateIC(xim,
6289 		    XNInputStyle, input_style,
6290 		    XNClientWindow, x11_window,
6291 		    XNFocusWindow, gui.wid,
6292 		    XNPreeditAttributes, preedit_list,
6293 		    XNStatusAttributes, status_list,
6294 		    NULL);
6295     XFree(status_list);
6296     XFree(preedit_list);
6297     if (xic != NULL)
6298     {
6299 	if (input_style & XIMStatusArea)
6300 	{
6301 	    xim_set_status_area();
6302 	    status_area_enabled = TRUE;
6303 	}
6304 	else
6305 	    gui_set_shellsize(FALSE, FALSE, RESIZE_BOTH);
6306     }
6307     else
6308     {
6309 	if (!is_not_a_term())
6310 	    emsg(_(e_xim));
6311 	XCloseIM(xim);
6312 	return FALSE;
6313     }
6314 
6315     return TRUE;
6316 }
6317 
6318 #  endif /* FEAT_GUI_X11 */
6319 
6320 /*
6321  * Get IM status.  When IM is on, return TRUE.  Else return FALSE.
6322  * FIXME: This doesn't work correctly: Having focus doesn't always mean XIM is
6323  * active, when not having focus XIM may still be active (e.g., when using a
6324  * tear-off menu item).
6325  */
6326     int
6327 im_get_status(void)
6328 {
6329 #  ifdef FEAT_EVAL
6330     if (USE_IMSTATUSFUNC)
6331 	return call_imstatusfunc();
6332 #  endif
6333     return xim_has_focus;
6334 }
6335 
6336 # endif /* !FEAT_GUI_GTK */
6337 
6338 # if !defined(FEAT_GUI_GTK) || defined(PROTO)
6339 /*
6340  * Set up the status area.
6341  *
6342  * This should use a separate Widget, but that seems not possible, because
6343  * preedit_area and status_area should be set to the same window as for the
6344  * text input.  Unfortunately this means the status area pollutes the text
6345  * window...
6346  */
6347     void
6348 xim_set_status_area(void)
6349 {
6350     XVaNestedList preedit_list = 0, status_list = 0, list = 0;
6351     XRectangle pre_area, status_area;
6352 
6353     if (xic == NULL)
6354 	return;
6355 
6356     if (input_style & XIMStatusArea)
6357     {
6358 	if (input_style & XIMPreeditArea)
6359 	{
6360 	    XRectangle *needed_rect;
6361 
6362 	    /* to get status_area width */
6363 	    status_list = XVaCreateNestedList(0, XNAreaNeeded,
6364 					      &needed_rect, NULL);
6365 	    XGetICValues(xic, XNStatusAttributes, status_list, NULL);
6366 	    XFree(status_list);
6367 
6368 	    status_area.width = needed_rect->width;
6369 	}
6370 	else
6371 	    status_area.width = gui.char_width * Columns;
6372 
6373 	status_area.x = 0;
6374 	status_area.y = gui.char_height * Rows + gui.border_offset;
6375 	if (gui.which_scrollbars[SBAR_BOTTOM])
6376 	    status_area.y += gui.scrollbar_height;
6377 #ifdef FEAT_MENU
6378 	if (gui.menu_is_active)
6379 	    status_area.y += gui.menu_height;
6380 #endif
6381 	status_area.height = gui.char_height;
6382 	status_list = XVaCreateNestedList(0, XNArea, &status_area, NULL);
6383     }
6384     else
6385     {
6386 	status_area.x = 0;
6387 	status_area.y = gui.char_height * Rows + gui.border_offset;
6388 	if (gui.which_scrollbars[SBAR_BOTTOM])
6389 	    status_area.y += gui.scrollbar_height;
6390 #ifdef FEAT_MENU
6391 	if (gui.menu_is_active)
6392 	    status_area.y += gui.menu_height;
6393 #endif
6394 	status_area.width = 0;
6395 	status_area.height = gui.char_height;
6396     }
6397 
6398     if (input_style & XIMPreeditArea)   /* off-the-spot */
6399     {
6400 	pre_area.x = status_area.x + status_area.width;
6401 	pre_area.y = gui.char_height * Rows + gui.border_offset;
6402 	pre_area.width = gui.char_width * Columns - pre_area.x;
6403 	if (gui.which_scrollbars[SBAR_BOTTOM])
6404 	    pre_area.y += gui.scrollbar_height;
6405 #ifdef FEAT_MENU
6406 	if (gui.menu_is_active)
6407 	    pre_area.y += gui.menu_height;
6408 #endif
6409 	pre_area.height = gui.char_height;
6410 	preedit_list = XVaCreateNestedList(0, XNArea, &pre_area, NULL);
6411     }
6412     else if (input_style & XIMPreeditPosition)   /* over-the-spot */
6413     {
6414 	pre_area.x = 0;
6415 	pre_area.y = 0;
6416 	pre_area.height = gui.char_height * Rows;
6417 	pre_area.width = gui.char_width * Columns;
6418 	preedit_list = XVaCreateNestedList(0, XNArea, &pre_area, NULL);
6419     }
6420 
6421     if (preedit_list && status_list)
6422 	list = XVaCreateNestedList(0, XNPreeditAttributes, preedit_list,
6423 				   XNStatusAttributes, status_list, NULL);
6424     else if (preedit_list)
6425 	list = XVaCreateNestedList(0, XNPreeditAttributes, preedit_list,
6426 				   NULL);
6427     else if (status_list)
6428 	list = XVaCreateNestedList(0, XNStatusAttributes, status_list,
6429 				   NULL);
6430     else
6431 	list = NULL;
6432 
6433     if (list)
6434     {
6435 	XSetICValues(xic, XNVaNestedList, list, NULL);
6436 	XFree(list);
6437     }
6438     if (status_list)
6439 	XFree(status_list);
6440     if (preedit_list)
6441 	XFree(preedit_list);
6442 }
6443 
6444     int
6445 xim_get_status_area_height(void)
6446 {
6447     if (status_area_enabled)
6448 	return gui.char_height;
6449     return 0;
6450 }
6451 # endif
6452 
6453 #else /* !defined(FEAT_XIM) */
6454 
6455 # ifdef IME_WITHOUT_XIM
6456 static int im_was_set_active = FALSE;
6457 
6458     int
6459 im_get_status(void)
6460 {
6461 #  if defined(FEAT_EVAL)
6462     if (USE_IMSTATUSFUNC)
6463 	return call_imstatusfunc();
6464 #  endif
6465     return im_was_set_active;
6466 }
6467 
6468     void
6469 im_set_active(int active_arg)
6470 {
6471 #  if defined(FEAT_EVAL)
6472     int	    active = !p_imdisable && active_arg;
6473 
6474     if (USE_IMACTIVATEFUNC && active != im_get_status())
6475     {
6476 	call_imactivatefunc(active);
6477 	im_was_set_active = active;
6478     }
6479 #  endif
6480 }
6481 
6482 #  ifdef FEAT_GUI
6483     void
6484 im_set_position(int row UNUSED, int col UNUSED)
6485 {
6486 }
6487 #  endif
6488 # endif
6489 
6490 #endif /* FEAT_XIM */
6491 
6492 
6493 /*
6494  * Setup "vcp" for conversion from "from" to "to".
6495  * The names must have been made canonical with enc_canonize().
6496  * vcp->vc_type must have been initialized to CONV_NONE.
6497  * Note: cannot be used for conversion from/to ucs-2 and ucs-4 (will use utf-8
6498  * instead).
6499  * Afterwards invoke with "from" and "to" equal to NULL to cleanup.
6500  * Return FAIL when conversion is not supported, OK otherwise.
6501  */
6502     int
6503 convert_setup(vimconv_T *vcp, char_u *from, char_u *to)
6504 {
6505     return convert_setup_ext(vcp, from, TRUE, to, TRUE);
6506 }
6507 
6508 /*
6509  * As convert_setup(), but only when from_unicode_is_utf8 is TRUE will all
6510  * "from" unicode charsets be considered utf-8.  Same for "to".
6511  */
6512     int
6513 convert_setup_ext(
6514     vimconv_T	*vcp,
6515     char_u	*from,
6516     int		from_unicode_is_utf8,
6517     char_u	*to,
6518     int		to_unicode_is_utf8)
6519 {
6520     int		from_prop;
6521     int		to_prop;
6522     int		from_is_utf8;
6523     int		to_is_utf8;
6524 
6525     /* Reset to no conversion. */
6526 #ifdef USE_ICONV
6527     if (vcp->vc_type == CONV_ICONV && vcp->vc_fd != (iconv_t)-1)
6528 	iconv_close(vcp->vc_fd);
6529 #endif
6530     vcp->vc_type = CONV_NONE;
6531     vcp->vc_factor = 1;
6532     vcp->vc_fail = FALSE;
6533 
6534     /* No conversion when one of the names is empty or they are equal. */
6535     if (from == NULL || *from == NUL || to == NULL || *to == NUL
6536 						     || STRCMP(from, to) == 0)
6537 	return OK;
6538 
6539     from_prop = enc_canon_props(from);
6540     to_prop = enc_canon_props(to);
6541     if (from_unicode_is_utf8)
6542 	from_is_utf8 = from_prop & ENC_UNICODE;
6543     else
6544 	from_is_utf8 = from_prop == ENC_UNICODE;
6545     if (to_unicode_is_utf8)
6546 	to_is_utf8 = to_prop & ENC_UNICODE;
6547     else
6548 	to_is_utf8 = to_prop == ENC_UNICODE;
6549 
6550     if ((from_prop & ENC_LATIN1) && to_is_utf8)
6551     {
6552 	/* Internal latin1 -> utf-8 conversion. */
6553 	vcp->vc_type = CONV_TO_UTF8;
6554 	vcp->vc_factor = 2;	/* up to twice as long */
6555     }
6556     else if ((from_prop & ENC_LATIN9) && to_is_utf8)
6557     {
6558 	/* Internal latin9 -> utf-8 conversion. */
6559 	vcp->vc_type = CONV_9_TO_UTF8;
6560 	vcp->vc_factor = 3;	/* up to three as long (euro sign) */
6561     }
6562     else if (from_is_utf8 && (to_prop & ENC_LATIN1))
6563     {
6564 	/* Internal utf-8 -> latin1 conversion. */
6565 	vcp->vc_type = CONV_TO_LATIN1;
6566     }
6567     else if (from_is_utf8 && (to_prop & ENC_LATIN9))
6568     {
6569 	/* Internal utf-8 -> latin9 conversion. */
6570 	vcp->vc_type = CONV_TO_LATIN9;
6571     }
6572 #ifdef MSWIN
6573     /* Win32-specific codepage <-> codepage conversion without iconv. */
6574     else if ((from_is_utf8 || encname2codepage(from) > 0)
6575 	    && (to_is_utf8 || encname2codepage(to) > 0))
6576     {
6577 	vcp->vc_type = CONV_CODEPAGE;
6578 	vcp->vc_factor = 2;	/* up to twice as long */
6579 	vcp->vc_cpfrom = from_is_utf8 ? 0 : encname2codepage(from);
6580 	vcp->vc_cpto = to_is_utf8 ? 0 : encname2codepage(to);
6581     }
6582 #endif
6583 #ifdef MACOS_CONVERT
6584     else if ((from_prop & ENC_MACROMAN) && (to_prop & ENC_LATIN1))
6585     {
6586 	vcp->vc_type = CONV_MAC_LATIN1;
6587     }
6588     else if ((from_prop & ENC_MACROMAN) && to_is_utf8)
6589     {
6590 	vcp->vc_type = CONV_MAC_UTF8;
6591 	vcp->vc_factor = 2;	/* up to twice as long */
6592     }
6593     else if ((from_prop & ENC_LATIN1) && (to_prop & ENC_MACROMAN))
6594     {
6595 	vcp->vc_type = CONV_LATIN1_MAC;
6596     }
6597     else if (from_is_utf8 && (to_prop & ENC_MACROMAN))
6598     {
6599 	vcp->vc_type = CONV_UTF8_MAC;
6600     }
6601 #endif
6602 #ifdef USE_ICONV
6603     else
6604     {
6605 	/* Use iconv() for conversion. */
6606 	vcp->vc_fd = (iconv_t)my_iconv_open(
6607 		to_is_utf8 ? (char_u *)"utf-8" : to,
6608 		from_is_utf8 ? (char_u *)"utf-8" : from);
6609 	if (vcp->vc_fd != (iconv_t)-1)
6610 	{
6611 	    vcp->vc_type = CONV_ICONV;
6612 	    vcp->vc_factor = 4;	/* could be longer too... */
6613 	}
6614     }
6615 #endif
6616     if (vcp->vc_type == CONV_NONE)
6617 	return FAIL;
6618 
6619     return OK;
6620 }
6621 
6622 #if defined(FEAT_GUI) || defined(AMIGA) || defined(MSWIN) \
6623 	|| defined(PROTO)
6624 /*
6625  * Do conversion on typed input characters in-place.
6626  * The input and output are not NUL terminated!
6627  * Returns the length after conversion.
6628  */
6629     int
6630 convert_input(char_u *ptr, int len, int maxlen)
6631 {
6632     return convert_input_safe(ptr, len, maxlen, NULL, NULL);
6633 }
6634 #endif
6635 
6636 /*
6637  * Like convert_input(), but when there is an incomplete byte sequence at the
6638  * end return that as an allocated string in "restp" and set "*restlenp" to
6639  * the length.  If "restp" is NULL it is not used.
6640  */
6641     int
6642 convert_input_safe(
6643     char_u	*ptr,
6644     int		len,
6645     int		maxlen,
6646     char_u	**restp,
6647     int		*restlenp)
6648 {
6649     char_u	*d;
6650     int		dlen = len;
6651     int		unconvertlen = 0;
6652 
6653     d = string_convert_ext(&input_conv, ptr, &dlen,
6654 					restp == NULL ? NULL : &unconvertlen);
6655     if (d != NULL)
6656     {
6657 	if (dlen <= maxlen)
6658 	{
6659 	    if (unconvertlen > 0)
6660 	    {
6661 		/* Move the unconverted characters to allocated memory. */
6662 		*restp = alloc(unconvertlen);
6663 		if (*restp != NULL)
6664 		    mch_memmove(*restp, ptr + len - unconvertlen, unconvertlen);
6665 		*restlenp = unconvertlen;
6666 	    }
6667 	    mch_memmove(ptr, d, dlen);
6668 	}
6669 	else
6670 	    /* result is too long, keep the unconverted text (the caller must
6671 	     * have done something wrong!) */
6672 	    dlen = len;
6673 	vim_free(d);
6674     }
6675     return dlen;
6676 }
6677 
6678 /*
6679  * Convert text "ptr[*lenp]" according to "vcp".
6680  * Returns the result in allocated memory and sets "*lenp".
6681  * When "lenp" is NULL, use NUL terminated strings.
6682  * Illegal chars are often changed to "?", unless vcp->vc_fail is set.
6683  * When something goes wrong, NULL is returned and "*lenp" is unchanged.
6684  */
6685     char_u *
6686 string_convert(
6687     vimconv_T	*vcp,
6688     char_u	*ptr,
6689     int		*lenp)
6690 {
6691     return string_convert_ext(vcp, ptr, lenp, NULL);
6692 }
6693 
6694 /*
6695  * Like string_convert(), but when "unconvlenp" is not NULL and there are is
6696  * an incomplete sequence at the end it is not converted and "*unconvlenp" is
6697  * set to the number of remaining bytes.
6698  */
6699     char_u *
6700 string_convert_ext(
6701     vimconv_T	*vcp,
6702     char_u	*ptr,
6703     int		*lenp,
6704     int		*unconvlenp)
6705 {
6706     char_u	*retval = NULL;
6707     char_u	*d;
6708     int		len;
6709     int		i;
6710     int		l;
6711     int		c;
6712 
6713     if (lenp == NULL)
6714 	len = (int)STRLEN(ptr);
6715     else
6716 	len = *lenp;
6717     if (len == 0)
6718 	return vim_strsave((char_u *)"");
6719 
6720     switch (vcp->vc_type)
6721     {
6722 	case CONV_TO_UTF8:	/* latin1 to utf-8 conversion */
6723 	    retval = alloc(len * 2 + 1);
6724 	    if (retval == NULL)
6725 		break;
6726 	    d = retval;
6727 	    for (i = 0; i < len; ++i)
6728 	    {
6729 		c = ptr[i];
6730 		if (c < 0x80)
6731 		    *d++ = c;
6732 		else
6733 		{
6734 		    *d++ = 0xc0 + ((unsigned)c >> 6);
6735 		    *d++ = 0x80 + (c & 0x3f);
6736 		}
6737 	    }
6738 	    *d = NUL;
6739 	    if (lenp != NULL)
6740 		*lenp = (int)(d - retval);
6741 	    break;
6742 
6743 	case CONV_9_TO_UTF8:	/* latin9 to utf-8 conversion */
6744 	    retval = alloc(len * 3 + 1);
6745 	    if (retval == NULL)
6746 		break;
6747 	    d = retval;
6748 	    for (i = 0; i < len; ++i)
6749 	    {
6750 		c = ptr[i];
6751 		switch (c)
6752 		{
6753 		    case 0xa4: c = 0x20ac; break;   /* euro */
6754 		    case 0xa6: c = 0x0160; break;   /* S hat */
6755 		    case 0xa8: c = 0x0161; break;   /* S -hat */
6756 		    case 0xb4: c = 0x017d; break;   /* Z hat */
6757 		    case 0xb8: c = 0x017e; break;   /* Z -hat */
6758 		    case 0xbc: c = 0x0152; break;   /* OE */
6759 		    case 0xbd: c = 0x0153; break;   /* oe */
6760 		    case 0xbe: c = 0x0178; break;   /* Y */
6761 		}
6762 		d += utf_char2bytes(c, d);
6763 	    }
6764 	    *d = NUL;
6765 	    if (lenp != NULL)
6766 		*lenp = (int)(d - retval);
6767 	    break;
6768 
6769 	case CONV_TO_LATIN1:	/* utf-8 to latin1 conversion */
6770 	case CONV_TO_LATIN9:	/* utf-8 to latin9 conversion */
6771 	    retval = alloc(len + 1);
6772 	    if (retval == NULL)
6773 		break;
6774 	    d = retval;
6775 	    for (i = 0; i < len; ++i)
6776 	    {
6777 		l = utf_ptr2len_len(ptr + i, len - i);
6778 		if (l == 0)
6779 		    *d++ = NUL;
6780 		else if (l == 1)
6781 		{
6782 		    int l_w = utf8len_tab_zero[ptr[i]];
6783 
6784 		    if (l_w == 0)
6785 		    {
6786 			/* Illegal utf-8 byte cannot be converted */
6787 			vim_free(retval);
6788 			return NULL;
6789 		    }
6790 		    if (unconvlenp != NULL && l_w > len - i)
6791 		    {
6792 			/* Incomplete sequence at the end. */
6793 			*unconvlenp = len - i;
6794 			break;
6795 		    }
6796 		    *d++ = ptr[i];
6797 		}
6798 		else
6799 		{
6800 		    c = utf_ptr2char(ptr + i);
6801 		    if (vcp->vc_type == CONV_TO_LATIN9)
6802 			switch (c)
6803 			{
6804 			    case 0x20ac: c = 0xa4; break;   /* euro */
6805 			    case 0x0160: c = 0xa6; break;   /* S hat */
6806 			    case 0x0161: c = 0xa8; break;   /* S -hat */
6807 			    case 0x017d: c = 0xb4; break;   /* Z hat */
6808 			    case 0x017e: c = 0xb8; break;   /* Z -hat */
6809 			    case 0x0152: c = 0xbc; break;   /* OE */
6810 			    case 0x0153: c = 0xbd; break;   /* oe */
6811 			    case 0x0178: c = 0xbe; break;   /* Y */
6812 			    case 0xa4:
6813 			    case 0xa6:
6814 			    case 0xa8:
6815 			    case 0xb4:
6816 			    case 0xb8:
6817 			    case 0xbc:
6818 			    case 0xbd:
6819 			    case 0xbe: c = 0x100; break; /* not in latin9 */
6820 			}
6821 		    if (!utf_iscomposing(c))	/* skip composing chars */
6822 		    {
6823 			if (c < 0x100)
6824 			    *d++ = c;
6825 			else if (vcp->vc_fail)
6826 			{
6827 			    vim_free(retval);
6828 			    return NULL;
6829 			}
6830 			else
6831 			{
6832 			    *d++ = 0xbf;
6833 			    if (utf_char2cells(c) > 1)
6834 				*d++ = '?';
6835 			}
6836 		    }
6837 		    i += l - 1;
6838 		}
6839 	    }
6840 	    *d = NUL;
6841 	    if (lenp != NULL)
6842 		*lenp = (int)(d - retval);
6843 	    break;
6844 
6845 # ifdef MACOS_CONVERT
6846 	case CONV_MAC_LATIN1:
6847 	    retval = mac_string_convert(ptr, len, lenp, vcp->vc_fail,
6848 					'm', 'l', unconvlenp);
6849 	    break;
6850 
6851 	case CONV_LATIN1_MAC:
6852 	    retval = mac_string_convert(ptr, len, lenp, vcp->vc_fail,
6853 					'l', 'm', unconvlenp);
6854 	    break;
6855 
6856 	case CONV_MAC_UTF8:
6857 	    retval = mac_string_convert(ptr, len, lenp, vcp->vc_fail,
6858 					'm', 'u', unconvlenp);
6859 	    break;
6860 
6861 	case CONV_UTF8_MAC:
6862 	    retval = mac_string_convert(ptr, len, lenp, vcp->vc_fail,
6863 					'u', 'm', unconvlenp);
6864 	    break;
6865 # endif
6866 
6867 # ifdef USE_ICONV
6868 	case CONV_ICONV:	/* conversion with output_conv.vc_fd */
6869 	    retval = iconv_string(vcp, ptr, len, unconvlenp, lenp);
6870 	    break;
6871 # endif
6872 # ifdef MSWIN
6873 	case CONV_CODEPAGE:		/* codepage -> codepage */
6874 	{
6875 	    int		retlen;
6876 	    int		tmp_len;
6877 	    short_u	*tmp;
6878 
6879 	    /* 1. codepage/UTF-8  ->  ucs-2. */
6880 	    if (vcp->vc_cpfrom == 0)
6881 		tmp_len = utf8_to_utf16(ptr, len, NULL, NULL);
6882 	    else
6883 	    {
6884 		tmp_len = MultiByteToWideChar(vcp->vc_cpfrom,
6885 					unconvlenp ? MB_ERR_INVALID_CHARS : 0,
6886 					(char *)ptr, len, 0, 0);
6887 		if (tmp_len == 0
6888 			&& GetLastError() == ERROR_NO_UNICODE_TRANSLATION)
6889 		{
6890 		    if (lenp != NULL)
6891 			*lenp = 0;
6892 		    if (unconvlenp != NULL)
6893 			*unconvlenp = len;
6894 		    retval = alloc(1);
6895 		    if (retval)
6896 			retval[0] = NUL;
6897 		    return retval;
6898 		}
6899 	    }
6900 	    tmp = (short_u *)alloc(sizeof(short_u) * tmp_len);
6901 	    if (tmp == NULL)
6902 		break;
6903 	    if (vcp->vc_cpfrom == 0)
6904 		utf8_to_utf16(ptr, len, tmp, unconvlenp);
6905 	    else
6906 		MultiByteToWideChar(vcp->vc_cpfrom, 0,
6907 			(char *)ptr, len, tmp, tmp_len);
6908 
6909 	    /* 2. ucs-2  ->  codepage/UTF-8. */
6910 	    if (vcp->vc_cpto == 0)
6911 		retlen = utf16_to_utf8(tmp, tmp_len, NULL);
6912 	    else
6913 		retlen = WideCharToMultiByte(vcp->vc_cpto, 0,
6914 						    tmp, tmp_len, 0, 0, 0, 0);
6915 	    retval = alloc(retlen + 1);
6916 	    if (retval != NULL)
6917 	    {
6918 		if (vcp->vc_cpto == 0)
6919 		    utf16_to_utf8(tmp, tmp_len, retval);
6920 		else
6921 		    WideCharToMultiByte(vcp->vc_cpto, 0,
6922 					  tmp, tmp_len,
6923 					  (char *)retval, retlen, 0, 0);
6924 		retval[retlen] = NUL;
6925 		if (lenp != NULL)
6926 		    *lenp = retlen;
6927 	    }
6928 	    vim_free(tmp);
6929 	    break;
6930 	}
6931 # endif
6932     }
6933 
6934     return retval;
6935 }
6936