xref: /vim-8.2.3635/src/mbyte.c (revision fc65cabb)
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(WIN3264) || 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 WIN3264
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 #if defined(FEAT_MBYTE) || defined(PROTO)
140 
141 static int enc_canon_search(char_u *name);
142 static int dbcs_char2len(int c);
143 static int dbcs_char2bytes(int c, char_u *buf);
144 static int dbcs_ptr2len(char_u *p);
145 static int dbcs_ptr2len_len(char_u *p, int size);
146 static int utf_ptr2cells_len(char_u *p, int size);
147 static int dbcs_char2cells(int c);
148 static int dbcs_ptr2cells_len(char_u *p, int size);
149 static int dbcs_ptr2char(char_u *p);
150 static int utf_safe_read_char_adv(char_u **s, size_t *n);
151 
152 /*
153  * Lookup table to quickly get the length in bytes of a UTF-8 character from
154  * the first byte of a UTF-8 string.
155  * Bytes which are illegal when used as the first byte have a 1.
156  * The NUL byte has length 1.
157  */
158 static char utf8len_tab[256] =
159 {
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     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,
163     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,
164     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,
165     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,
166     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,
167     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,
168 };
169 
170 /*
171  * Like utf8len_tab above, but using a zero for illegal lead bytes.
172  */
173 static char utf8len_tab_zero[256] =
174 {
175     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,
176     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,
177     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,
178     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,
179     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,
180     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,
181     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,
182     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,
183 };
184 
185 /*
186  * XIM often causes trouble.  Define XIM_DEBUG to get a log of XIM callbacks
187  * in the "xim.log" file.
188  */
189 /* #define XIM_DEBUG */
190 #ifdef XIM_DEBUG
191     static void
192 xim_log(char *s, ...)
193 {
194     va_list arglist;
195     static FILE *fd = NULL;
196 
197     if (fd == (FILE *)-1)
198 	return;
199     if (fd == NULL)
200     {
201 	fd = mch_fopen("xim.log", "w");
202 	if (fd == NULL)
203 	{
204 	    EMSG("Cannot open xim.log");
205 	    fd = (FILE *)-1;
206 	    return;
207 	}
208     }
209 
210     va_start(arglist, s);
211     vfprintf(fd, s, arglist);
212     va_end(arglist);
213 }
214 #endif
215 
216 #endif
217 
218 #if defined(FEAT_MBYTE) || defined(FEAT_POSTSCRIPT) || defined(PROTO)
219 /*
220  * Canonical encoding names and their properties.
221  * "iso-8859-n" is handled by enc_canonize() directly.
222  */
223 static struct
224 {   char *name;		int prop;		int codepage;}
225 enc_canon_table[] =
226 {
227 #define IDX_LATIN_1	0
228     {"latin1",		ENC_8BIT + ENC_LATIN1,	1252},
229 #define IDX_ISO_2	1
230     {"iso-8859-2",	ENC_8BIT,		0},
231 #define IDX_ISO_3	2
232     {"iso-8859-3",	ENC_8BIT,		0},
233 #define IDX_ISO_4	3
234     {"iso-8859-4",	ENC_8BIT,		0},
235 #define IDX_ISO_5	4
236     {"iso-8859-5",	ENC_8BIT,		0},
237 #define IDX_ISO_6	5
238     {"iso-8859-6",	ENC_8BIT,		0},
239 #define IDX_ISO_7	6
240     {"iso-8859-7",	ENC_8BIT,		0},
241 #define IDX_ISO_8	7
242     {"iso-8859-8",	ENC_8BIT,		0},
243 #define IDX_ISO_9	8
244     {"iso-8859-9",	ENC_8BIT,		0},
245 #define IDX_ISO_10	9
246     {"iso-8859-10",	ENC_8BIT,		0},
247 #define IDX_ISO_11	10
248     {"iso-8859-11",	ENC_8BIT,		0},
249 #define IDX_ISO_13	11
250     {"iso-8859-13",	ENC_8BIT,		0},
251 #define IDX_ISO_14	12
252     {"iso-8859-14",	ENC_8BIT,		0},
253 #define IDX_ISO_15	13
254     {"iso-8859-15",	ENC_8BIT + ENC_LATIN9,	0},
255 #define IDX_KOI8_R	14
256     {"koi8-r",		ENC_8BIT,		0},
257 #define IDX_KOI8_U	15
258     {"koi8-u",		ENC_8BIT,		0},
259 #define IDX_UTF8	16
260     {"utf-8",		ENC_UNICODE,		0},
261 #define IDX_UCS2	17
262     {"ucs-2",		ENC_UNICODE + ENC_ENDIAN_B + ENC_2BYTE, 0},
263 #define IDX_UCS2LE	18
264     {"ucs-2le",		ENC_UNICODE + ENC_ENDIAN_L + ENC_2BYTE, 0},
265 #define IDX_UTF16	19
266     {"utf-16",		ENC_UNICODE + ENC_ENDIAN_B + ENC_2WORD, 0},
267 #define IDX_UTF16LE	20
268     {"utf-16le",	ENC_UNICODE + ENC_ENDIAN_L + ENC_2WORD, 0},
269 #define IDX_UCS4	21
270     {"ucs-4",		ENC_UNICODE + ENC_ENDIAN_B + ENC_4BYTE, 0},
271 #define IDX_UCS4LE	22
272     {"ucs-4le",		ENC_UNICODE + ENC_ENDIAN_L + ENC_4BYTE, 0},
273 
274     /* For debugging DBCS encoding on Unix. */
275 #define IDX_DEBUG	23
276     {"debug",		ENC_DBCS,		DBCS_DEBUG},
277 #define IDX_EUC_JP	24
278     {"euc-jp",		ENC_DBCS,		DBCS_JPNU},
279 #define IDX_SJIS	25
280     {"sjis",		ENC_DBCS,		DBCS_JPN},
281 #define IDX_EUC_KR	26
282     {"euc-kr",		ENC_DBCS,		DBCS_KORU},
283 #define IDX_EUC_CN	27
284     {"euc-cn",		ENC_DBCS,		DBCS_CHSU},
285 #define IDX_EUC_TW	28
286     {"euc-tw",		ENC_DBCS,		DBCS_CHTU},
287 #define IDX_BIG5	29
288     {"big5",		ENC_DBCS,		DBCS_CHT},
289 
290     /* MS-DOS and MS-Windows codepages are included here, so that they can be
291      * used on Unix too.  Most of them are similar to ISO-8859 encodings, but
292      * not exactly the same. */
293 #define IDX_CP437	30
294     {"cp437",		ENC_8BIT,		437}, /* like iso-8859-1 */
295 #define IDX_CP737	31
296     {"cp737",		ENC_8BIT,		737}, /* like iso-8859-7 */
297 #define IDX_CP775	32
298     {"cp775",		ENC_8BIT,		775}, /* Baltic */
299 #define IDX_CP850	33
300     {"cp850",		ENC_8BIT,		850}, /* like iso-8859-4 */
301 #define IDX_CP852	34
302     {"cp852",		ENC_8BIT,		852}, /* like iso-8859-1 */
303 #define IDX_CP855	35
304     {"cp855",		ENC_8BIT,		855}, /* like iso-8859-2 */
305 #define IDX_CP857	36
306     {"cp857",		ENC_8BIT,		857}, /* like iso-8859-5 */
307 #define IDX_CP860	37
308     {"cp860",		ENC_8BIT,		860}, /* like iso-8859-9 */
309 #define IDX_CP861	38
310     {"cp861",		ENC_8BIT,		861}, /* like iso-8859-1 */
311 #define IDX_CP862	39
312     {"cp862",		ENC_8BIT,		862}, /* like iso-8859-1 */
313 #define IDX_CP863	40
314     {"cp863",		ENC_8BIT,		863}, /* like iso-8859-8 */
315 #define IDX_CP865	41
316     {"cp865",		ENC_8BIT,		865}, /* like iso-8859-1 */
317 #define IDX_CP866	42
318     {"cp866",		ENC_8BIT,		866}, /* like iso-8859-5 */
319 #define IDX_CP869	43
320     {"cp869",		ENC_8BIT,		869}, /* like iso-8859-7 */
321 #define IDX_CP874	44
322     {"cp874",		ENC_8BIT,		874}, /* Thai */
323 #define IDX_CP932	45
324     {"cp932",		ENC_DBCS,		DBCS_JPN},
325 #define IDX_CP936	46
326     {"cp936",		ENC_DBCS,		DBCS_CHS},
327 #define IDX_CP949	47
328     {"cp949",		ENC_DBCS,		DBCS_KOR},
329 #define IDX_CP950	48
330     {"cp950",		ENC_DBCS,		DBCS_CHT},
331 #define IDX_CP1250	49
332     {"cp1250",		ENC_8BIT,		1250}, /* Czech, Polish, etc. */
333 #define IDX_CP1251	50
334     {"cp1251",		ENC_8BIT,		1251}, /* Cyrillic */
335     /* cp1252 is considered to be equal to latin1 */
336 #define IDX_CP1253	51
337     {"cp1253",		ENC_8BIT,		1253}, /* Greek */
338 #define IDX_CP1254	52
339     {"cp1254",		ENC_8BIT,		1254}, /* Turkish */
340 #define IDX_CP1255	53
341     {"cp1255",		ENC_8BIT,		1255}, /* Hebrew */
342 #define IDX_CP1256	54
343     {"cp1256",		ENC_8BIT,		1256}, /* Arabic */
344 #define IDX_CP1257	55
345     {"cp1257",		ENC_8BIT,		1257}, /* Baltic */
346 #define IDX_CP1258	56
347     {"cp1258",		ENC_8BIT,		1258}, /* Vietnamese */
348 
349 #define IDX_MACROMAN	57
350     {"macroman",	ENC_8BIT + ENC_MACROMAN, 0},	/* Mac OS */
351 #define IDX_DECMCS	58
352     {"dec-mcs",		ENC_8BIT,		0},	/* DEC MCS */
353 #define IDX_HPROMAN8	59
354     {"hp-roman8",	ENC_8BIT,		0},	/* HP Roman8 */
355 #define IDX_COUNT	60
356 };
357 
358 /*
359  * Aliases for encoding names.
360  */
361 static struct
362 {   char *name;		int canon;}
363 enc_alias_table[] =
364 {
365     {"ansi",		IDX_LATIN_1},
366     {"iso-8859-1",	IDX_LATIN_1},
367     {"latin2",		IDX_ISO_2},
368     {"latin3",		IDX_ISO_3},
369     {"latin4",		IDX_ISO_4},
370     {"cyrillic",	IDX_ISO_5},
371     {"arabic",		IDX_ISO_6},
372     {"greek",		IDX_ISO_7},
373 #ifdef WIN3264
374     {"hebrew",		IDX_CP1255},
375 #else
376     {"hebrew",		IDX_ISO_8},
377 #endif
378     {"latin5",		IDX_ISO_9},
379     {"turkish",		IDX_ISO_9}, /* ? */
380     {"latin6",		IDX_ISO_10},
381     {"nordic",		IDX_ISO_10}, /* ? */
382     {"thai",		IDX_ISO_11}, /* ? */
383     {"latin7",		IDX_ISO_13},
384     {"latin8",		IDX_ISO_14},
385     {"latin9",		IDX_ISO_15},
386     {"utf8",		IDX_UTF8},
387     {"unicode",		IDX_UCS2},
388     {"ucs2",		IDX_UCS2},
389     {"ucs2be",		IDX_UCS2},
390     {"ucs-2be",		IDX_UCS2},
391     {"ucs2le",		IDX_UCS2LE},
392     {"utf16",		IDX_UTF16},
393     {"utf16be",		IDX_UTF16},
394     {"utf-16be",	IDX_UTF16},
395     {"utf16le",		IDX_UTF16LE},
396     {"ucs4",		IDX_UCS4},
397     {"ucs4be",		IDX_UCS4},
398     {"ucs-4be",		IDX_UCS4},
399     {"ucs4le",		IDX_UCS4LE},
400     {"utf32",		IDX_UCS4},
401     {"utf-32",		IDX_UCS4},
402     {"utf32be",		IDX_UCS4},
403     {"utf-32be",	IDX_UCS4},
404     {"utf32le",		IDX_UCS4LE},
405     {"utf-32le",	IDX_UCS4LE},
406     {"932",		IDX_CP932},
407     {"949",		IDX_CP949},
408     {"936",		IDX_CP936},
409     {"gbk",		IDX_CP936},
410     {"950",		IDX_CP950},
411     {"eucjp",		IDX_EUC_JP},
412     {"unix-jis",	IDX_EUC_JP},
413     {"ujis",		IDX_EUC_JP},
414     {"shift-jis",	IDX_SJIS},
415     {"pck",		IDX_SJIS},	/* Sun: PCK */
416     {"euckr",		IDX_EUC_KR},
417     {"5601",		IDX_EUC_KR},	/* Sun: KS C 5601 */
418     {"euccn",		IDX_EUC_CN},
419     {"gb2312",		IDX_EUC_CN},
420     {"euctw",		IDX_EUC_TW},
421 #if defined(WIN3264) || defined(WIN32UNIX) || defined(MACOS_X)
422     {"japan",		IDX_CP932},
423     {"korea",		IDX_CP949},
424     {"prc",		IDX_CP936},
425     {"chinese",		IDX_CP936},
426     {"taiwan",		IDX_CP950},
427     {"big5",		IDX_CP950},
428 #else
429     {"japan",		IDX_EUC_JP},
430     {"korea",		IDX_EUC_KR},
431     {"prc",		IDX_EUC_CN},
432     {"chinese",		IDX_EUC_CN},
433     {"taiwan",		IDX_EUC_TW},
434     {"cp950",		IDX_BIG5},
435     {"950",		IDX_BIG5},
436 #endif
437     {"mac",		IDX_MACROMAN},
438     {"mac-roman",	IDX_MACROMAN},
439     {NULL,		0}
440 };
441 
442 #ifndef CP_UTF8
443 # define CP_UTF8 65001	/* magic number from winnls.h */
444 #endif
445 
446 /*
447  * Find encoding "name" in the list of canonical encoding names.
448  * Returns -1 if not found.
449  */
450     static int
451 enc_canon_search(char_u *name)
452 {
453     int		i;
454 
455     for (i = 0; i < IDX_COUNT; ++i)
456 	if (STRCMP(name, enc_canon_table[i].name) == 0)
457 	    return i;
458     return -1;
459 }
460 
461 #endif
462 
463 #if defined(FEAT_MBYTE) || defined(PROTO)
464 
465 /*
466  * Find canonical encoding "name" in the list and return its properties.
467  * Returns 0 if not found.
468  */
469     int
470 enc_canon_props(char_u *name)
471 {
472     int		i;
473 
474     i = enc_canon_search(name);
475     if (i >= 0)
476 	return enc_canon_table[i].prop;
477 #ifdef WIN3264
478     if (name[0] == 'c' && name[1] == 'p' && VIM_ISDIGIT(name[2]))
479     {
480 	CPINFO	cpinfo;
481 
482 	/* Get info on this codepage to find out what it is. */
483 	if (GetCPInfo(atoi((char *)name + 2), &cpinfo) != 0)
484 	{
485 	    if (cpinfo.MaxCharSize == 1) /* some single-byte encoding */
486 		return ENC_8BIT;
487 	    if (cpinfo.MaxCharSize == 2
488 		    && (cpinfo.LeadByte[0] != 0 || cpinfo.LeadByte[1] != 0))
489 		/* must be a DBCS encoding */
490 		return ENC_DBCS;
491 	}
492 	return 0;
493     }
494 #endif
495     if (STRNCMP(name, "2byte-", 6) == 0)
496 	return ENC_DBCS;
497     if (STRNCMP(name, "8bit-", 5) == 0 || STRNCMP(name, "iso-8859-", 9) == 0)
498 	return ENC_8BIT;
499     return 0;
500 }
501 
502 /*
503  * Set up for using multi-byte characters.
504  * Called in three cases:
505  * - by main() to initialize (p_enc == NULL)
506  * - by set_init_1() after 'encoding' was set to its default.
507  * - by do_set() when 'encoding' has been set.
508  * p_enc must have been passed through enc_canonize() already.
509  * Sets the "enc_unicode", "enc_utf8", "enc_dbcs" and "has_mbyte" flags.
510  * Fills mb_bytelen_tab[] and returns NULL when there are no problems.
511  * When there is something wrong: Returns an error message and doesn't change
512  * anything.
513  */
514     char_u *
515 mb_init(void)
516 {
517     int		i;
518     int		idx;
519     int		n;
520     int		enc_dbcs_new = 0;
521 #if defined(USE_ICONV) && !defined(WIN3264) && !defined(WIN32UNIX) \
522 	&& !defined(MACOS_CONVERT)
523 # define LEN_FROM_CONV
524     vimconv_T	vimconv;
525     char_u	*p;
526 #endif
527 
528     if (p_enc == NULL)
529     {
530 	/* Just starting up: set the whole table to one's. */
531 	for (i = 0; i < 256; ++i)
532 	    mb_bytelen_tab[i] = 1;
533 	input_conv.vc_type = CONV_NONE;
534 	input_conv.vc_factor = 1;
535 	output_conv.vc_type = CONV_NONE;
536 	return NULL;
537     }
538 
539 #ifdef WIN3264
540     if (p_enc[0] == 'c' && p_enc[1] == 'p' && VIM_ISDIGIT(p_enc[2]))
541     {
542 	CPINFO	cpinfo;
543 
544 	/* Get info on this codepage to find out what it is. */
545 	if (GetCPInfo(atoi((char *)p_enc + 2), &cpinfo) != 0)
546 	{
547 	    if (cpinfo.MaxCharSize == 1)
548 	    {
549 		/* some single-byte encoding */
550 		enc_unicode = 0;
551 		enc_utf8 = FALSE;
552 	    }
553 	    else if (cpinfo.MaxCharSize == 2
554 		    && (cpinfo.LeadByte[0] != 0 || cpinfo.LeadByte[1] != 0))
555 	    {
556 		/* must be a DBCS encoding, check below */
557 		enc_dbcs_new = atoi((char *)p_enc + 2);
558 	    }
559 	    else
560 		goto codepage_invalid;
561 	}
562 	else if (GetLastError() == ERROR_INVALID_PARAMETER)
563 	{
564 codepage_invalid:
565 	    return (char_u *)N_("E543: Not a valid codepage");
566 	}
567     }
568 #endif
569     else if (STRNCMP(p_enc, "8bit-", 5) == 0
570 	    || STRNCMP(p_enc, "iso-8859-", 9) == 0)
571     {
572 	/* Accept any "8bit-" or "iso-8859-" name. */
573 	enc_unicode = 0;
574 	enc_utf8 = FALSE;
575     }
576     else if (STRNCMP(p_enc, "2byte-", 6) == 0)
577     {
578 #ifdef WIN3264
579 	/* Windows: accept only valid codepage numbers, check below. */
580 	if (p_enc[6] != 'c' || p_enc[7] != 'p'
581 			      || (enc_dbcs_new = atoi((char *)p_enc + 8)) == 0)
582 	    return e_invarg;
583 #else
584 	/* Unix: accept any "2byte-" name, assume current locale. */
585 	enc_dbcs_new = DBCS_2BYTE;
586 #endif
587     }
588     else if ((idx = enc_canon_search(p_enc)) >= 0)
589     {
590 	i = enc_canon_table[idx].prop;
591 	if (i & ENC_UNICODE)
592 	{
593 	    /* Unicode */
594 	    enc_utf8 = TRUE;
595 	    if (i & (ENC_2BYTE | ENC_2WORD))
596 		enc_unicode = 2;
597 	    else if (i & ENC_4BYTE)
598 		enc_unicode = 4;
599 	    else
600 		enc_unicode = 0;
601 	}
602 	else if (i & ENC_DBCS)
603 	{
604 	    /* 2byte, handle below */
605 	    enc_dbcs_new = enc_canon_table[idx].codepage;
606 	}
607 	else
608 	{
609 	    /* Must be 8-bit. */
610 	    enc_unicode = 0;
611 	    enc_utf8 = FALSE;
612 	}
613     }
614     else    /* Don't know what encoding this is, reject it. */
615 	return e_invarg;
616 
617     if (enc_dbcs_new != 0)
618     {
619 #ifdef WIN3264
620 	/* Check if the DBCS code page is OK. */
621 	if (!IsValidCodePage(enc_dbcs_new))
622 	    goto codepage_invalid;
623 #endif
624 	enc_unicode = 0;
625 	enc_utf8 = FALSE;
626     }
627     enc_dbcs = enc_dbcs_new;
628     has_mbyte = (enc_dbcs != 0 || enc_utf8);
629 
630 #if defined(WIN3264) || defined(FEAT_CYGWIN_WIN32_CLIPBOARD)
631     enc_codepage = encname2codepage(p_enc);
632     enc_latin9 = (STRCMP(p_enc, "iso-8859-15") == 0);
633 #endif
634 
635     /* Detect an encoding that uses latin1 characters. */
636     enc_latin1like = (enc_utf8 || STRCMP(p_enc, "latin1") == 0
637 					|| STRCMP(p_enc, "iso-8859-15") == 0);
638 
639     /*
640      * Set the function pointers.
641      */
642     if (enc_utf8)
643     {
644 	mb_ptr2len = utfc_ptr2len;
645 	mb_ptr2len_len = utfc_ptr2len_len;
646 	mb_char2len = utf_char2len;
647 	mb_char2bytes = utf_char2bytes;
648 	mb_ptr2cells = utf_ptr2cells;
649 	mb_ptr2cells_len = utf_ptr2cells_len;
650 	mb_char2cells = utf_char2cells;
651 	mb_off2cells = utf_off2cells;
652 	mb_ptr2char = utf_ptr2char;
653 	mb_head_off = utf_head_off;
654     }
655     else if (enc_dbcs != 0)
656     {
657 	mb_ptr2len = dbcs_ptr2len;
658 	mb_ptr2len_len = dbcs_ptr2len_len;
659 	mb_char2len = dbcs_char2len;
660 	mb_char2bytes = dbcs_char2bytes;
661 	mb_ptr2cells = dbcs_ptr2cells;
662 	mb_ptr2cells_len = dbcs_ptr2cells_len;
663 	mb_char2cells = dbcs_char2cells;
664 	mb_off2cells = dbcs_off2cells;
665 	mb_ptr2char = dbcs_ptr2char;
666 	mb_head_off = dbcs_head_off;
667     }
668     else
669     {
670 	mb_ptr2len = latin_ptr2len;
671 	mb_ptr2len_len = latin_ptr2len_len;
672 	mb_char2len = latin_char2len;
673 	mb_char2bytes = latin_char2bytes;
674 	mb_ptr2cells = latin_ptr2cells;
675 	mb_ptr2cells_len = latin_ptr2cells_len;
676 	mb_char2cells = latin_char2cells;
677 	mb_off2cells = latin_off2cells;
678 	mb_ptr2char = latin_ptr2char;
679 	mb_head_off = latin_head_off;
680     }
681 
682     /*
683      * Fill the mb_bytelen_tab[] for MB_BYTE2LEN().
684      */
685 #ifdef LEN_FROM_CONV
686     /* When 'encoding' is different from the current locale mblen() won't
687      * work.  Use conversion to "utf-8" instead. */
688     vimconv.vc_type = CONV_NONE;
689     if (enc_dbcs)
690     {
691 	p = enc_locale();
692 	if (p == NULL || STRCMP(p, p_enc) != 0)
693 	{
694 	    convert_setup(&vimconv, p_enc, (char_u *)"utf-8");
695 	    vimconv.vc_fail = TRUE;
696 	}
697 	vim_free(p);
698     }
699 #endif
700 
701     for (i = 0; i < 256; ++i)
702     {
703 	/* Our own function to reliably check the length of UTF-8 characters,
704 	 * independent of mblen(). */
705 	if (enc_utf8)
706 	    n = utf8len_tab[i];
707 	else if (enc_dbcs == 0)
708 	    n = 1;
709 	else
710 	{
711 #if defined(WIN3264) || defined(WIN32UNIX)
712 	    /* enc_dbcs is set by setting 'fileencoding'.  It becomes a Windows
713 	     * CodePage identifier, which we can pass directly in to Windows
714 	     * API */
715 	    n = IsDBCSLeadByteEx(enc_dbcs, (WINBYTE)i) ? 2 : 1;
716 #else
717 # if defined(__amigaos4__) || defined(__ANDROID__) || \
718 				   !(defined(HAVE_MBLEN) || defined(X_LOCALE))
719 	    /*
720 	     * if mblen() is not available, character which MSB is turned on
721 	     * are treated as leading byte character. (note : This assumption
722 	     * is not always true.)
723 	     */
724 	    n = (i & 0x80) ? 2 : 1;
725 # else
726 	    char buf[MB_MAXBYTES + 1];
727 
728 	    if (i == NUL)	/* just in case mblen() can't handle "" */
729 		n = 1;
730 	    else
731 	    {
732 		buf[0] = i;
733 		buf[1] = 0;
734 #  ifdef LEN_FROM_CONV
735 		if (vimconv.vc_type != CONV_NONE)
736 		{
737 		    /*
738 		     * string_convert() should fail when converting the first
739 		     * byte of a double-byte character.
740 		     */
741 		    p = string_convert(&vimconv, (char_u *)buf, NULL);
742 		    if (p != NULL)
743 		    {
744 			vim_free(p);
745 			n = 1;
746 		    }
747 		    else
748 			n = 2;
749 		}
750 		else
751 #  endif
752 		{
753 		    /*
754 		     * mblen() should return -1 for invalid (means the leading
755 		     * multibyte) character.  However there are some platforms
756 		     * where mblen() returns 0 for invalid character.
757 		     * Therefore, following condition includes 0.
758 		     */
759 		    ignored = mblen(NULL, 0);	/* First reset the state. */
760 		    if (mblen(buf, (size_t)1) <= 0)
761 			n = 2;
762 		    else
763 			n = 1;
764 		}
765 	    }
766 # endif
767 #endif
768 	}
769 
770 	mb_bytelen_tab[i] = n;
771     }
772 
773 #ifdef LEN_FROM_CONV
774     convert_setup(&vimconv, NULL, NULL);
775 #endif
776 
777     /* The cell width depends on the type of multi-byte characters. */
778     (void)init_chartab();
779 
780     /* When enc_utf8 is set or reset, (de)allocate ScreenLinesUC[] */
781     screenalloc(FALSE);
782 
783     /* When using Unicode, set default for 'fileencodings'. */
784     if (enc_utf8 && !option_was_set((char_u *)"fencs"))
785 	set_string_option_direct((char_u *)"fencs", -1,
786 		       (char_u *)"ucs-bom,utf-8,default,latin1", OPT_FREE, 0);
787 
788 #if defined(HAVE_BIND_TEXTDOMAIN_CODESET) && defined(FEAT_GETTEXT)
789     /* GNU gettext 0.10.37 supports this feature: set the codeset used for
790      * translated messages independently from the current locale. */
791     (void)bind_textdomain_codeset(VIMPACKAGE,
792 					  enc_utf8 ? "utf-8" : (char *)p_enc);
793 #endif
794 
795 #ifdef WIN32
796     /* When changing 'encoding' while starting up, then convert the command
797      * line arguments from the active codepage to 'encoding'. */
798     if (starting != 0)
799 	fix_arg_enc();
800 #endif
801 
802     /* Fire an autocommand to let people do custom font setup. This must be
803      * after Vim has been setup for the new encoding. */
804     apply_autocmds(EVENT_ENCODINGCHANGED, NULL, (char_u *)"", FALSE, curbuf);
805 
806 #ifdef FEAT_SPELL
807     /* Need to reload spell dictionaries */
808     spell_reload();
809 #endif
810 
811     return NULL;
812 }
813 
814 /*
815  * Return the size of the BOM for the current buffer:
816  * 0 - no BOM
817  * 2 - UCS-2 or UTF-16 BOM
818  * 4 - UCS-4 BOM
819  * 3 - UTF-8 BOM
820  */
821     int
822 bomb_size(void)
823 {
824     int n = 0;
825 
826     if (curbuf->b_p_bomb && !curbuf->b_p_bin)
827     {
828 	if (*curbuf->b_p_fenc == NUL)
829 	{
830 	    if (enc_utf8)
831 	    {
832 		if (enc_unicode != 0)
833 		    n = enc_unicode;
834 		else
835 		    n = 3;
836 	    }
837 	}
838 	else if (STRCMP(curbuf->b_p_fenc, "utf-8") == 0)
839 	    n = 3;
840 	else if (STRNCMP(curbuf->b_p_fenc, "ucs-2", 5) == 0
841 		|| STRNCMP(curbuf->b_p_fenc, "utf-16", 6) == 0)
842 	    n = 2;
843 	else if (STRNCMP(curbuf->b_p_fenc, "ucs-4", 5) == 0)
844 	    n = 4;
845     }
846     return n;
847 }
848 
849 /*
850  * Remove all BOM from "s" by moving remaining text.
851  */
852     void
853 remove_bom(char_u *s)
854 {
855     if (enc_utf8)
856     {
857 	char_u *p = s;
858 
859 	while ((p = vim_strbyte(p, 0xef)) != NULL)
860 	{
861 	    if (p[1] == 0xbb && p[2] == 0xbf)
862 		STRMOVE(p, p + 3);
863 	    else
864 		++p;
865 	}
866     }
867 }
868 
869 /*
870  * Get class of pointer:
871  * 0 for blank or NUL
872  * 1 for punctuation
873  * 2 for an (ASCII) word character
874  * >2 for other word characters
875  */
876     int
877 mb_get_class(char_u *p)
878 {
879     return mb_get_class_buf(p, curbuf);
880 }
881 
882     int
883 mb_get_class_buf(char_u *p, buf_T *buf)
884 {
885     if (MB_BYTE2LEN(p[0]) == 1)
886     {
887 	if (p[0] == NUL || VIM_ISWHITE(p[0]))
888 	    return 0;
889 	if (vim_iswordc_buf(p[0], buf))
890 	    return 2;
891 	return 1;
892     }
893     if (enc_dbcs != 0 && p[0] != NUL && p[1] != NUL)
894 	return dbcs_class(p[0], p[1]);
895     if (enc_utf8)
896 	return utf_class_buf(utf_ptr2char(p), buf);
897     return 0;
898 }
899 
900 /*
901  * Get class of a double-byte character.  This always returns 3 or bigger.
902  * TODO: Should return 1 for punctuation.
903  */
904     int
905 dbcs_class(unsigned lead, unsigned trail)
906 {
907     switch (enc_dbcs)
908     {
909 	/* please add classify routine for your language in here */
910 
911 	case DBCS_JPNU:	/* ? */
912 	case DBCS_JPN:
913 	    {
914 		/* JIS code classification */
915 		unsigned char lb = lead;
916 		unsigned char tb = trail;
917 
918 		/* convert process code to JIS */
919 # if defined(WIN3264) || defined(WIN32UNIX) || defined(MACOS_X)
920 		/* process code is SJIS */
921 		if (lb <= 0x9f)
922 		    lb = (lb - 0x81) * 2 + 0x21;
923 		else
924 		    lb = (lb - 0xc1) * 2 + 0x21;
925 		if (tb <= 0x7e)
926 		    tb -= 0x1f;
927 		else if (tb <= 0x9e)
928 		    tb -= 0x20;
929 		else
930 		{
931 		    tb -= 0x7e;
932 		    lb += 1;
933 		}
934 # else
935 		/*
936 		 * XXX: Code page identification can not use with all
937 		 *	    system! So, some other encoding information
938 		 *	    will be needed.
939 		 *	    In japanese: SJIS,EUC,UNICODE,(JIS)
940 		 *	    Note that JIS-code system don't use as
941 		 *	    process code in most system because it uses
942 		 *	    escape sequences(JIS is context depend encoding).
943 		 */
944 		/* assume process code is JAPANESE-EUC */
945 		lb &= 0x7f;
946 		tb &= 0x7f;
947 # endif
948 		/* exceptions */
949 		switch (lb << 8 | tb)
950 		{
951 		    case 0x2121: /* ZENKAKU space */
952 			return 0;
953 		    case 0x2122: /* TOU-TEN (Japanese comma) */
954 		    case 0x2123: /* KU-TEN (Japanese period) */
955 		    case 0x2124: /* ZENKAKU comma */
956 		    case 0x2125: /* ZENKAKU period */
957 			return 1;
958 		    case 0x213c: /* prolongedsound handled as KATAKANA */
959 			return 13;
960 		}
961 		/* sieved by KU code */
962 		switch (lb)
963 		{
964 		    case 0x21:
965 		    case 0x22:
966 			/* special symbols */
967 			return 10;
968 		    case 0x23:
969 			/* alpha-numeric */
970 			return 11;
971 		    case 0x24:
972 			/* hiragana */
973 			return 12;
974 		    case 0x25:
975 			/* katakana */
976 			return 13;
977 		    case 0x26:
978 			/* greek */
979 			return 14;
980 		    case 0x27:
981 			/* russian */
982 			return 15;
983 		    case 0x28:
984 			/* lines */
985 			return 16;
986 		    default:
987 			/* kanji */
988 			return 17;
989 		}
990 	    }
991 
992 	case DBCS_KORU:	/* ? */
993 	case DBCS_KOR:
994 	    {
995 		/* KS code classification */
996 		unsigned char c1 = lead;
997 		unsigned char c2 = trail;
998 
999 		/*
1000 		 * 20 : Hangul
1001 		 * 21 : Hanja
1002 		 * 22 : Symbols
1003 		 * 23 : Alpha-numeric/Roman Letter (Full width)
1004 		 * 24 : Hangul Letter(Alphabet)
1005 		 * 25 : Roman Numeral/Greek Letter
1006 		 * 26 : Box Drawings
1007 		 * 27 : Unit Symbols
1008 		 * 28 : Circled/Parenthesized Letter
1009 		 * 29 : Hiragana/Katakana
1010 		 * 30 : Cyrillic Letter
1011 		 */
1012 
1013 		if (c1 >= 0xB0 && c1 <= 0xC8)
1014 		    /* Hangul */
1015 		    return 20;
1016 #if defined(WIN3264) || defined(WIN32UNIX)
1017 		else if (c1 <= 0xA0 || c2 <= 0xA0)
1018 		    /* Extended Hangul Region : MS UHC(Unified Hangul Code) */
1019 		    /* c1: 0x81-0xA0 with c2: 0x41-0x5A, 0x61-0x7A, 0x81-0xFE
1020 		     * c1: 0xA1-0xC6 with c2: 0x41-0x5A, 0x61-0x7A, 0x81-0xA0
1021 		     */
1022 		    return 20;
1023 #endif
1024 
1025 		else if (c1 >= 0xCA && c1 <= 0xFD)
1026 		    /* Hanja */
1027 		    return 21;
1028 		else switch (c1)
1029 		{
1030 		    case 0xA1:
1031 		    case 0xA2:
1032 			/* Symbols */
1033 			return 22;
1034 		    case 0xA3:
1035 			/* Alpha-numeric */
1036 			return 23;
1037 		    case 0xA4:
1038 			/* Hangul Letter(Alphabet) */
1039 			return 24;
1040 		    case 0xA5:
1041 			/* Roman Numeral/Greek Letter */
1042 			return 25;
1043 		    case 0xA6:
1044 			/* Box Drawings */
1045 			return 26;
1046 		    case 0xA7:
1047 			/* Unit Symbols */
1048 			return 27;
1049 		    case 0xA8:
1050 		    case 0xA9:
1051 			if (c2 <= 0xAF)
1052 			    return 25;  /* Roman Letter */
1053 			else if (c2 >= 0xF6)
1054 			    return 22;  /* Symbols */
1055 			else
1056 			    /* Circled/Parenthesized Letter */
1057 			    return 28;
1058 		    case 0xAA:
1059 		    case 0xAB:
1060 			/* Hiragana/Katakana */
1061 			return 29;
1062 		    case 0xAC:
1063 			/* Cyrillic Letter */
1064 			return 30;
1065 		}
1066 	    }
1067 	default:
1068 	    break;
1069     }
1070     return 3;
1071 }
1072 
1073 /*
1074  * mb_char2len() function pointer.
1075  * Return length in bytes of character "c".
1076  * Returns 1 for a single-byte character.
1077  */
1078     int
1079 latin_char2len(int c UNUSED)
1080 {
1081     return 1;
1082 }
1083 
1084     static int
1085 dbcs_char2len(
1086     int		c)
1087 {
1088     if (c >= 0x100)
1089 	return 2;
1090     return 1;
1091 }
1092 
1093 /*
1094  * mb_char2bytes() function pointer.
1095  * Convert a character to its bytes.
1096  * Returns the length in bytes.
1097  */
1098     int
1099 latin_char2bytes(int c, char_u *buf)
1100 {
1101     buf[0] = c;
1102     return 1;
1103 }
1104 
1105     static int
1106 dbcs_char2bytes(int c, char_u *buf)
1107 {
1108     if (c >= 0x100)
1109     {
1110 	buf[0] = (unsigned)c >> 8;
1111 	buf[1] = c;
1112 	/* Never use a NUL byte, it causes lots of trouble.  It's an invalid
1113 	 * character anyway. */
1114 	if (buf[1] == NUL)
1115 	    buf[1] = '\n';
1116 	return 2;
1117     }
1118     buf[0] = c;
1119     return 1;
1120 }
1121 
1122 /*
1123  * mb_ptr2len() function pointer.
1124  * Get byte length of character at "*p" but stop at a NUL.
1125  * For UTF-8 this includes following composing characters.
1126  * Returns 0 when *p is NUL.
1127  */
1128     int
1129 latin_ptr2len(char_u *p)
1130 {
1131  return MB_BYTE2LEN(*p);
1132 }
1133 
1134     static int
1135 dbcs_ptr2len(
1136     char_u	*p)
1137 {
1138     int		len;
1139 
1140     /* Check if second byte is not missing. */
1141     len = MB_BYTE2LEN(*p);
1142     if (len == 2 && p[1] == NUL)
1143 	len = 1;
1144     return len;
1145 }
1146 
1147 /*
1148  * mb_ptr2len_len() function pointer.
1149  * Like mb_ptr2len(), but limit to read "size" bytes.
1150  * Returns 0 for an empty string.
1151  * Returns 1 for an illegal char or an incomplete byte sequence.
1152  */
1153     int
1154 latin_ptr2len_len(char_u *p, int size)
1155 {
1156     if (size < 1 || *p == NUL)
1157 	return 0;
1158     return 1;
1159 }
1160 
1161     static int
1162 dbcs_ptr2len_len(char_u *p, int size)
1163 {
1164     int		len;
1165 
1166     if (size < 1 || *p == NUL)
1167 	return 0;
1168     if (size == 1)
1169 	return 1;
1170     /* Check that second byte is not missing. */
1171     len = MB_BYTE2LEN(*p);
1172     if (len == 2 && p[1] == NUL)
1173 	len = 1;
1174     return len;
1175 }
1176 
1177 struct interval
1178 {
1179     long first;
1180     long last;
1181 };
1182 
1183 /*
1184  * Return TRUE if "c" is in "table[size / sizeof(struct interval)]".
1185  */
1186     static int
1187 intable(struct interval *table, size_t size, int c)
1188 {
1189     int mid, bot, top;
1190 
1191     /* first quick check for Latin1 etc. characters */
1192     if (c < table[0].first)
1193 	return FALSE;
1194 
1195     /* binary search in table */
1196     bot = 0;
1197     top = (int)(size / sizeof(struct interval) - 1);
1198     while (top >= bot)
1199     {
1200 	mid = (bot + top) / 2;
1201 	if (table[mid].last < c)
1202 	    bot = mid + 1;
1203 	else if (table[mid].first > c)
1204 	    top = mid - 1;
1205 	else
1206 	    return TRUE;
1207     }
1208     return FALSE;
1209 }
1210 
1211 /* Sorted list of non-overlapping intervals of East Asian Ambiguous
1212  * characters, generated with ../runtime/tools/unicode.vim. */
1213 static struct interval ambiguous[] =
1214 {
1215     {0x00a1, 0x00a1},
1216     {0x00a4, 0x00a4},
1217     {0x00a7, 0x00a8},
1218     {0x00aa, 0x00aa},
1219     {0x00ad, 0x00ae},
1220     {0x00b0, 0x00b4},
1221     {0x00b6, 0x00ba},
1222     {0x00bc, 0x00bf},
1223     {0x00c6, 0x00c6},
1224     {0x00d0, 0x00d0},
1225     {0x00d7, 0x00d8},
1226     {0x00de, 0x00e1},
1227     {0x00e6, 0x00e6},
1228     {0x00e8, 0x00ea},
1229     {0x00ec, 0x00ed},
1230     {0x00f0, 0x00f0},
1231     {0x00f2, 0x00f3},
1232     {0x00f7, 0x00fa},
1233     {0x00fc, 0x00fc},
1234     {0x00fe, 0x00fe},
1235     {0x0101, 0x0101},
1236     {0x0111, 0x0111},
1237     {0x0113, 0x0113},
1238     {0x011b, 0x011b},
1239     {0x0126, 0x0127},
1240     {0x012b, 0x012b},
1241     {0x0131, 0x0133},
1242     {0x0138, 0x0138},
1243     {0x013f, 0x0142},
1244     {0x0144, 0x0144},
1245     {0x0148, 0x014b},
1246     {0x014d, 0x014d},
1247     {0x0152, 0x0153},
1248     {0x0166, 0x0167},
1249     {0x016b, 0x016b},
1250     {0x01ce, 0x01ce},
1251     {0x01d0, 0x01d0},
1252     {0x01d2, 0x01d2},
1253     {0x01d4, 0x01d4},
1254     {0x01d6, 0x01d6},
1255     {0x01d8, 0x01d8},
1256     {0x01da, 0x01da},
1257     {0x01dc, 0x01dc},
1258     {0x0251, 0x0251},
1259     {0x0261, 0x0261},
1260     {0x02c4, 0x02c4},
1261     {0x02c7, 0x02c7},
1262     {0x02c9, 0x02cb},
1263     {0x02cd, 0x02cd},
1264     {0x02d0, 0x02d0},
1265     {0x02d8, 0x02db},
1266     {0x02dd, 0x02dd},
1267     {0x02df, 0x02df},
1268     {0x0300, 0x036f},
1269     {0x0391, 0x03a1},
1270     {0x03a3, 0x03a9},
1271     {0x03b1, 0x03c1},
1272     {0x03c3, 0x03c9},
1273     {0x0401, 0x0401},
1274     {0x0410, 0x044f},
1275     {0x0451, 0x0451},
1276     {0x2010, 0x2010},
1277     {0x2013, 0x2016},
1278     {0x2018, 0x2019},
1279     {0x201c, 0x201d},
1280     {0x2020, 0x2022},
1281     {0x2024, 0x2027},
1282     {0x2030, 0x2030},
1283     {0x2032, 0x2033},
1284     {0x2035, 0x2035},
1285     {0x203b, 0x203b},
1286     {0x203e, 0x203e},
1287     {0x2074, 0x2074},
1288     {0x207f, 0x207f},
1289     {0x2081, 0x2084},
1290     {0x20ac, 0x20ac},
1291     {0x2103, 0x2103},
1292     {0x2105, 0x2105},
1293     {0x2109, 0x2109},
1294     {0x2113, 0x2113},
1295     {0x2116, 0x2116},
1296     {0x2121, 0x2122},
1297     {0x2126, 0x2126},
1298     {0x212b, 0x212b},
1299     {0x2153, 0x2154},
1300     {0x215b, 0x215e},
1301     {0x2160, 0x216b},
1302     {0x2170, 0x2179},
1303     {0x2189, 0x2189},
1304     {0x2190, 0x2199},
1305     {0x21b8, 0x21b9},
1306     {0x21d2, 0x21d2},
1307     {0x21d4, 0x21d4},
1308     {0x21e7, 0x21e7},
1309     {0x2200, 0x2200},
1310     {0x2202, 0x2203},
1311     {0x2207, 0x2208},
1312     {0x220b, 0x220b},
1313     {0x220f, 0x220f},
1314     {0x2211, 0x2211},
1315     {0x2215, 0x2215},
1316     {0x221a, 0x221a},
1317     {0x221d, 0x2220},
1318     {0x2223, 0x2223},
1319     {0x2225, 0x2225},
1320     {0x2227, 0x222c},
1321     {0x222e, 0x222e},
1322     {0x2234, 0x2237},
1323     {0x223c, 0x223d},
1324     {0x2248, 0x2248},
1325     {0x224c, 0x224c},
1326     {0x2252, 0x2252},
1327     {0x2260, 0x2261},
1328     {0x2264, 0x2267},
1329     {0x226a, 0x226b},
1330     {0x226e, 0x226f},
1331     {0x2282, 0x2283},
1332     {0x2286, 0x2287},
1333     {0x2295, 0x2295},
1334     {0x2299, 0x2299},
1335     {0x22a5, 0x22a5},
1336     {0x22bf, 0x22bf},
1337     {0x2312, 0x2312},
1338     {0x2460, 0x24e9},
1339     {0x24eb, 0x254b},
1340     {0x2550, 0x2573},
1341     {0x2580, 0x258f},
1342     {0x2592, 0x2595},
1343     {0x25a0, 0x25a1},
1344     {0x25a3, 0x25a9},
1345     {0x25b2, 0x25b3},
1346     {0x25b6, 0x25b7},
1347     {0x25bc, 0x25bd},
1348     {0x25c0, 0x25c1},
1349     {0x25c6, 0x25c8},
1350     {0x25cb, 0x25cb},
1351     {0x25ce, 0x25d1},
1352     {0x25e2, 0x25e5},
1353     {0x25ef, 0x25ef},
1354     {0x2605, 0x2606},
1355     {0x2609, 0x2609},
1356     {0x260e, 0x260f},
1357     {0x261c, 0x261c},
1358     {0x261e, 0x261e},
1359     {0x2640, 0x2640},
1360     {0x2642, 0x2642},
1361     {0x2660, 0x2661},
1362     {0x2663, 0x2665},
1363     {0x2667, 0x266a},
1364     {0x266c, 0x266d},
1365     {0x266f, 0x266f},
1366     {0x269e, 0x269f},
1367     {0x26bf, 0x26bf},
1368     {0x26c6, 0x26cd},
1369     {0x26cf, 0x26d3},
1370     {0x26d5, 0x26e1},
1371     {0x26e3, 0x26e3},
1372     {0x26e8, 0x26e9},
1373     {0x26eb, 0x26f1},
1374     {0x26f4, 0x26f4},
1375     {0x26f6, 0x26f9},
1376     {0x26fb, 0x26fc},
1377     {0x26fe, 0x26ff},
1378     {0x273d, 0x273d},
1379     {0x2776, 0x277f},
1380     {0x2b56, 0x2b59},
1381     {0x3248, 0x324f},
1382     {0xe000, 0xf8ff},
1383     {0xfe00, 0xfe0f},
1384     {0xfffd, 0xfffd},
1385     {0x1f100, 0x1f10a},
1386     {0x1f110, 0x1f12d},
1387     {0x1f130, 0x1f169},
1388     {0x1f170, 0x1f18d},
1389     {0x1f18f, 0x1f190},
1390     {0x1f19b, 0x1f1ac},
1391     {0xe0100, 0xe01ef},
1392     {0xf0000, 0xffffd},
1393     {0x100000, 0x10fffd}
1394 };
1395 
1396 #if defined(FEAT_TERMINAL) || defined(PROTO)
1397 /*
1398  * utf_char2cells() with different argument type for libvterm.
1399  */
1400     int
1401 utf_uint2cells(UINT32_T c)
1402 {
1403     if (c >= 0x100 && utf_iscomposing((int)c))
1404 	return 0;
1405     return utf_char2cells((int)c);
1406 }
1407 #endif
1408 
1409 /*
1410  * For UTF-8 character "c" return 2 for a double-width character, 1 for others.
1411  * Returns 4 or 6 for an unprintable character.
1412  * Is only correct for characters >= 0x80.
1413  * When p_ambw is "double", return 2 for a character with East Asian Width
1414  * class 'A'(mbiguous).
1415  */
1416     int
1417 utf_char2cells(int c)
1418 {
1419     /* Sorted list of non-overlapping intervals of East Asian double width
1420      * characters, generated with ../runtime/tools/unicode.vim. */
1421     static struct interval doublewidth[] =
1422     {
1423 	{0x1100, 0x115f},
1424 	{0x231a, 0x231b},
1425 	{0x2329, 0x232a},
1426 	{0x23e9, 0x23ec},
1427 	{0x23f0, 0x23f0},
1428 	{0x23f3, 0x23f3},
1429 	{0x25fd, 0x25fe},
1430 	{0x2614, 0x2615},
1431 	{0x2648, 0x2653},
1432 	{0x267f, 0x267f},
1433 	{0x2693, 0x2693},
1434 	{0x26a1, 0x26a1},
1435 	{0x26aa, 0x26ab},
1436 	{0x26bd, 0x26be},
1437 	{0x26c4, 0x26c5},
1438 	{0x26ce, 0x26ce},
1439 	{0x26d4, 0x26d4},
1440 	{0x26ea, 0x26ea},
1441 	{0x26f2, 0x26f3},
1442 	{0x26f5, 0x26f5},
1443 	{0x26fa, 0x26fa},
1444 	{0x26fd, 0x26fd},
1445 	{0x2705, 0x2705},
1446 	{0x270a, 0x270b},
1447 	{0x2728, 0x2728},
1448 	{0x274c, 0x274c},
1449 	{0x274e, 0x274e},
1450 	{0x2753, 0x2755},
1451 	{0x2757, 0x2757},
1452 	{0x2795, 0x2797},
1453 	{0x27b0, 0x27b0},
1454 	{0x27bf, 0x27bf},
1455 	{0x2b1b, 0x2b1c},
1456 	{0x2b50, 0x2b50},
1457 	{0x2b55, 0x2b55},
1458 	{0x2e80, 0x2e99},
1459 	{0x2e9b, 0x2ef3},
1460 	{0x2f00, 0x2fd5},
1461 	{0x2ff0, 0x2ffb},
1462 	{0x3000, 0x303e},
1463 	{0x3041, 0x3096},
1464 	{0x3099, 0x30ff},
1465 	{0x3105, 0x312f},
1466 	{0x3131, 0x318e},
1467 	{0x3190, 0x31ba},
1468 	{0x31c0, 0x31e3},
1469 	{0x31f0, 0x321e},
1470 	{0x3220, 0x3247},
1471 	{0x3250, 0x32fe},
1472 	{0x3300, 0x4dbf},
1473 	{0x4e00, 0xa48c},
1474 	{0xa490, 0xa4c6},
1475 	{0xa960, 0xa97c},
1476 	{0xac00, 0xd7a3},
1477 	{0xf900, 0xfaff},
1478 	{0xfe10, 0xfe19},
1479 	{0xfe30, 0xfe52},
1480 	{0xfe54, 0xfe66},
1481 	{0xfe68, 0xfe6b},
1482 	{0xff01, 0xff60},
1483 	{0xffe0, 0xffe6},
1484 	{0x16fe0, 0x16fe1},
1485 	{0x17000, 0x187f1},
1486 	{0x18800, 0x18af2},
1487 	{0x1b000, 0x1b11e},
1488 	{0x1b170, 0x1b2fb},
1489 	{0x1f004, 0x1f004},
1490 	{0x1f0cf, 0x1f0cf},
1491 	{0x1f18e, 0x1f18e},
1492 	{0x1f191, 0x1f19a},
1493 	{0x1f200, 0x1f202},
1494 	{0x1f210, 0x1f23b},
1495 	{0x1f240, 0x1f248},
1496 	{0x1f250, 0x1f251},
1497 	{0x1f260, 0x1f265},
1498 	{0x1f300, 0x1f320},
1499 	{0x1f32d, 0x1f335},
1500 	{0x1f337, 0x1f37c},
1501 	{0x1f37e, 0x1f393},
1502 	{0x1f3a0, 0x1f3ca},
1503 	{0x1f3cf, 0x1f3d3},
1504 	{0x1f3e0, 0x1f3f0},
1505 	{0x1f3f4, 0x1f3f4},
1506 	{0x1f3f8, 0x1f43e},
1507 	{0x1f440, 0x1f440},
1508 	{0x1f442, 0x1f4fc},
1509 	{0x1f4ff, 0x1f53d},
1510 	{0x1f54b, 0x1f54e},
1511 	{0x1f550, 0x1f567},
1512 	{0x1f57a, 0x1f57a},
1513 	{0x1f595, 0x1f596},
1514 	{0x1f5a4, 0x1f5a4},
1515 	{0x1f5fb, 0x1f64f},
1516 	{0x1f680, 0x1f6c5},
1517 	{0x1f6cc, 0x1f6cc},
1518 	{0x1f6d0, 0x1f6d2},
1519 	{0x1f6eb, 0x1f6ec},
1520 	{0x1f6f4, 0x1f6f9},
1521 	{0x1f910, 0x1f93e},
1522 	{0x1f940, 0x1f970},
1523 	{0x1f973, 0x1f976},
1524 	{0x1f97a, 0x1f97a},
1525 	{0x1f97c, 0x1f9a2},
1526 	{0x1f9b0, 0x1f9b9},
1527 	{0x1f9c0, 0x1f9c2},
1528 	{0x1f9d0, 0x1f9ff},
1529 	{0x20000, 0x2fffd},
1530 	{0x30000, 0x3fffd}
1531     };
1532 
1533     /* Sorted list of non-overlapping intervals of Emoji characters that don't
1534      * have ambiguous or double width,
1535      * based on http://unicode.org/emoji/charts/emoji-list.html */
1536     static struct interval emoji_width[] =
1537     {
1538 	{0x1f1e6, 0x1f1ff},
1539 	{0x1f321, 0x1f321},
1540 	{0x1f324, 0x1f32c},
1541 	{0x1f336, 0x1f336},
1542 	{0x1f37d, 0x1f37d},
1543 	{0x1f396, 0x1f397},
1544 	{0x1f399, 0x1f39b},
1545 	{0x1f39e, 0x1f39f},
1546 	{0x1f3cb, 0x1f3ce},
1547 	{0x1f3d4, 0x1f3df},
1548 	{0x1f3f3, 0x1f3f5},
1549 	{0x1f3f7, 0x1f3f7},
1550 	{0x1f43f, 0x1f43f},
1551 	{0x1f441, 0x1f441},
1552 	{0x1f4fd, 0x1f4fd},
1553 	{0x1f549, 0x1f54a},
1554 	{0x1f56f, 0x1f570},
1555 	{0x1f573, 0x1f579},
1556 	{0x1f587, 0x1f587},
1557 	{0x1f58a, 0x1f58d},
1558 	{0x1f590, 0x1f590},
1559 	{0x1f5a5, 0x1f5a5},
1560 	{0x1f5a8, 0x1f5a8},
1561 	{0x1f5b1, 0x1f5b2},
1562 	{0x1f5bc, 0x1f5bc},
1563 	{0x1f5c2, 0x1f5c4},
1564 	{0x1f5d1, 0x1f5d3},
1565 	{0x1f5dc, 0x1f5de},
1566 	{0x1f5e1, 0x1f5e1},
1567 	{0x1f5e3, 0x1f5e3},
1568 	{0x1f5e8, 0x1f5e8},
1569 	{0x1f5ef, 0x1f5ef},
1570 	{0x1f5f3, 0x1f5f3},
1571 	{0x1f5fa, 0x1f5fa},
1572 	{0x1f6cb, 0x1f6cf},
1573 	{0x1f6e0, 0x1f6e5},
1574 	{0x1f6e9, 0x1f6e9},
1575 	{0x1f6f0, 0x1f6f0},
1576 	{0x1f6f3, 0x1f6f3}
1577     };
1578 
1579     if (c >= 0x100)
1580     {
1581 #ifdef USE_WCHAR_FUNCTIONS
1582 	/*
1583 	 * Assume the library function wcwidth() works better than our own
1584 	 * stuff.  It should return 1 for ambiguous width chars!
1585 	 */
1586 	int	n = wcwidth(c);
1587 
1588 	if (n < 0)
1589 	    return 6;		/* unprintable, displays <xxxx> */
1590 	if (n > 1)
1591 	    return n;
1592 #else
1593 	if (!utf_printable(c))
1594 	    return 6;		/* unprintable, displays <xxxx> */
1595 	if (intable(doublewidth, sizeof(doublewidth), c))
1596 	    return 2;
1597 #endif
1598 	if (p_emoji && intable(emoji_width, sizeof(emoji_width), c))
1599 	    return 2;
1600     }
1601 
1602     /* Characters below 0x100 are influenced by 'isprint' option */
1603     else if (c >= 0x80 && !vim_isprintc(c))
1604 	return 4;		/* unprintable, displays <xx> */
1605 
1606     if (c >= 0x80 && *p_ambw == 'd' && intable(ambiguous, sizeof(ambiguous), c))
1607 	return 2;
1608 
1609     return 1;
1610 }
1611 
1612 /*
1613  * mb_ptr2cells() function pointer.
1614  * Return the number of display cells character at "*p" occupies.
1615  * This doesn't take care of unprintable characters, use ptr2cells() for that.
1616  */
1617     int
1618 latin_ptr2cells(char_u *p UNUSED)
1619 {
1620     return 1;
1621 }
1622 
1623     int
1624 utf_ptr2cells(
1625     char_u	*p)
1626 {
1627     int		c;
1628 
1629     /* Need to convert to a wide character. */
1630     if (*p >= 0x80)
1631     {
1632 	c = utf_ptr2char(p);
1633 	/* An illegal byte is displayed as <xx>. */
1634 	if (utf_ptr2len(p) == 1 || c == NUL)
1635 	    return 4;
1636 	/* If the char is ASCII it must be an overlong sequence. */
1637 	if (c < 0x80)
1638 	    return char2cells(c);
1639 	return utf_char2cells(c);
1640     }
1641     return 1;
1642 }
1643 
1644     int
1645 dbcs_ptr2cells(char_u *p)
1646 {
1647     /* Number of cells is equal to number of bytes, except for euc-jp when
1648      * the first byte is 0x8e. */
1649     if (enc_dbcs == DBCS_JPNU && *p == 0x8e)
1650 	return 1;
1651     return MB_BYTE2LEN(*p);
1652 }
1653 
1654 /*
1655  * mb_ptr2cells_len() function pointer.
1656  * Like mb_ptr2cells(), but limit string length to "size".
1657  * For an empty string or truncated character returns 1.
1658  */
1659     int
1660 latin_ptr2cells_len(char_u *p UNUSED, int size UNUSED)
1661 {
1662     return 1;
1663 }
1664 
1665     static int
1666 utf_ptr2cells_len(char_u *p, int size)
1667 {
1668     int		c;
1669 
1670     /* Need to convert to a wide character. */
1671     if (size > 0 && *p >= 0x80)
1672     {
1673 	if (utf_ptr2len_len(p, size) < utf8len_tab[*p])
1674 	    return 1;  /* truncated */
1675 	c = utf_ptr2char(p);
1676 	/* An illegal byte is displayed as <xx>. */
1677 	if (utf_ptr2len(p) == 1 || c == NUL)
1678 	    return 4;
1679 	/* If the char is ASCII it must be an overlong sequence. */
1680 	if (c < 0x80)
1681 	    return char2cells(c);
1682 	return utf_char2cells(c);
1683     }
1684     return 1;
1685 }
1686 
1687     static int
1688 dbcs_ptr2cells_len(char_u *p, int size)
1689 {
1690     /* Number of cells is equal to number of bytes, except for euc-jp when
1691      * the first byte is 0x8e. */
1692     if (size <= 1 || (enc_dbcs == DBCS_JPNU && *p == 0x8e))
1693 	return 1;
1694     return MB_BYTE2LEN(*p);
1695 }
1696 
1697 /*
1698  * mb_char2cells() function pointer.
1699  * Return the number of display cells character "c" occupies.
1700  * Only takes care of multi-byte chars, not "^C" and such.
1701  */
1702     int
1703 latin_char2cells(int c UNUSED)
1704 {
1705     return 1;
1706 }
1707 
1708     static int
1709 dbcs_char2cells(int c)
1710 {
1711     /* Number of cells is equal to number of bytes, except for euc-jp when
1712      * the first byte is 0x8e. */
1713     if (enc_dbcs == DBCS_JPNU && ((unsigned)c >> 8) == 0x8e)
1714 	return 1;
1715     /* use the first byte */
1716     return MB_BYTE2LEN((unsigned)c >> 8);
1717 }
1718 
1719 /*
1720  * Return the number of cells occupied by string "p".
1721  * Stop at a NUL character.  When "len" >= 0 stop at character "p[len]".
1722  */
1723     int
1724 mb_string2cells(char_u *p, int len)
1725 {
1726     int i;
1727     int clen = 0;
1728 
1729     for (i = 0; (len < 0 || i < len) && p[i] != NUL; i += (*mb_ptr2len)(p + i))
1730 	clen += (*mb_ptr2cells)(p + i);
1731     return clen;
1732 }
1733 
1734 /*
1735  * mb_off2cells() function pointer.
1736  * Return number of display cells for char at ScreenLines[off].
1737  * We make sure that the offset used is less than "max_off".
1738  */
1739     int
1740 latin_off2cells(unsigned off UNUSED, unsigned max_off UNUSED)
1741 {
1742     return 1;
1743 }
1744 
1745     int
1746 dbcs_off2cells(unsigned off, unsigned max_off)
1747 {
1748     /* never check beyond end of the line */
1749     if (off >= max_off)
1750 	return 1;
1751 
1752     /* Number of cells is equal to number of bytes, except for euc-jp when
1753      * the first byte is 0x8e. */
1754     if (enc_dbcs == DBCS_JPNU && ScreenLines[off] == 0x8e)
1755 	return 1;
1756     return MB_BYTE2LEN(ScreenLines[off]);
1757 }
1758 
1759     int
1760 utf_off2cells(unsigned off, unsigned max_off)
1761 {
1762     return (off + 1 < max_off && ScreenLines[off + 1] == 0) ? 2 : 1;
1763 }
1764 
1765 /*
1766  * mb_ptr2char() function pointer.
1767  * Convert a byte sequence into a character.
1768  */
1769     int
1770 latin_ptr2char(char_u *p)
1771 {
1772     return *p;
1773 }
1774 
1775     static int
1776 dbcs_ptr2char(char_u *p)
1777 {
1778     if (MB_BYTE2LEN(*p) > 1 && p[1] != NUL)
1779 	return (p[0] << 8) + p[1];
1780     return *p;
1781 }
1782 
1783 /*
1784  * Convert a UTF-8 byte sequence to a wide character.
1785  * If the sequence is illegal or truncated by a NUL the first byte is
1786  * returned.
1787  * For an overlong sequence this may return zero.
1788  * Does not include composing characters, of course.
1789  */
1790     int
1791 utf_ptr2char(char_u *p)
1792 {
1793     int		len;
1794 
1795     if (p[0] < 0x80)	/* be quick for ASCII */
1796 	return p[0];
1797 
1798     len = utf8len_tab_zero[p[0]];
1799     if (len > 1 && (p[1] & 0xc0) == 0x80)
1800     {
1801 	if (len == 2)
1802 	    return ((p[0] & 0x1f) << 6) + (p[1] & 0x3f);
1803 	if ((p[2] & 0xc0) == 0x80)
1804 	{
1805 	    if (len == 3)
1806 		return ((p[0] & 0x0f) << 12) + ((p[1] & 0x3f) << 6)
1807 		    + (p[2] & 0x3f);
1808 	    if ((p[3] & 0xc0) == 0x80)
1809 	    {
1810 		if (len == 4)
1811 		    return ((p[0] & 0x07) << 18) + ((p[1] & 0x3f) << 12)
1812 			+ ((p[2] & 0x3f) << 6) + (p[3] & 0x3f);
1813 		if ((p[4] & 0xc0) == 0x80)
1814 		{
1815 		    if (len == 5)
1816 			return ((p[0] & 0x03) << 24) + ((p[1] & 0x3f) << 18)
1817 			    + ((p[2] & 0x3f) << 12) + ((p[3] & 0x3f) << 6)
1818 			    + (p[4] & 0x3f);
1819 		    if ((p[5] & 0xc0) == 0x80 && len == 6)
1820 			return ((p[0] & 0x01) << 30) + ((p[1] & 0x3f) << 24)
1821 			    + ((p[2] & 0x3f) << 18) + ((p[3] & 0x3f) << 12)
1822 			    + ((p[4] & 0x3f) << 6) + (p[5] & 0x3f);
1823 		}
1824 	    }
1825 	}
1826     }
1827     /* Illegal value, just return the first byte */
1828     return p[0];
1829 }
1830 
1831 /*
1832  * Convert a UTF-8 byte sequence to a wide character.
1833  * String is assumed to be terminated by NUL or after "n" bytes, whichever
1834  * comes first.
1835  * The function is safe in the sense that it never accesses memory beyond the
1836  * first "n" bytes of "s".
1837  *
1838  * On success, returns decoded codepoint, advances "s" to the beginning of
1839  * next character and decreases "n" accordingly.
1840  *
1841  * If end of string was reached, returns 0 and, if "n" > 0, advances "s" past
1842  * NUL byte.
1843  *
1844  * If byte sequence is illegal or incomplete, returns -1 and does not advance
1845  * "s".
1846  */
1847     static int
1848 utf_safe_read_char_adv(char_u **s, size_t *n)
1849 {
1850     int		c, k;
1851 
1852     if (*n == 0) /* end of buffer */
1853 	return 0;
1854 
1855     k = utf8len_tab_zero[**s];
1856 
1857     if (k == 1)
1858     {
1859 	/* ASCII character or NUL */
1860 	(*n)--;
1861 	return *(*s)++;
1862     }
1863 
1864     if ((size_t)k <= *n)
1865     {
1866 	/* We have a multibyte sequence and it isn't truncated by buffer
1867 	 * limits so utf_ptr2char() is safe to use. Or the first byte is
1868 	 * illegal (k=0), and it's also safe to use utf_ptr2char(). */
1869 	c = utf_ptr2char(*s);
1870 
1871 	/* On failure, utf_ptr2char() returns the first byte, so here we
1872 	 * check equality with the first byte. The only non-ASCII character
1873 	 * which equals the first byte of its own UTF-8 representation is
1874 	 * U+00C3 (UTF-8: 0xC3 0x83), so need to check that special case too.
1875 	 * It's safe even if n=1, else we would have k=2 > n. */
1876 	if (c != (int)(**s) || (c == 0xC3 && (*s)[1] == 0x83))
1877 	{
1878 	    /* byte sequence was successfully decoded */
1879 	    *s += k;
1880 	    *n -= k;
1881 	    return c;
1882 	}
1883     }
1884 
1885     /* byte sequence is incomplete or illegal */
1886     return -1;
1887 }
1888 
1889 /*
1890  * Get character at **pp and advance *pp to the next character.
1891  * Note: composing characters are skipped!
1892  */
1893     int
1894 mb_ptr2char_adv(char_u **pp)
1895 {
1896     int		c;
1897 
1898     c = (*mb_ptr2char)(*pp);
1899     *pp += (*mb_ptr2len)(*pp);
1900     return c;
1901 }
1902 
1903 /*
1904  * Get character at **pp and advance *pp to the next character.
1905  * Note: composing characters are returned as separate characters.
1906  */
1907     int
1908 mb_cptr2char_adv(char_u **pp)
1909 {
1910     int		c;
1911 
1912     c = (*mb_ptr2char)(*pp);
1913     if (enc_utf8)
1914 	*pp += utf_ptr2len(*pp);
1915     else
1916 	*pp += (*mb_ptr2len)(*pp);
1917     return c;
1918 }
1919 
1920 #if defined(FEAT_ARABIC) || defined(PROTO)
1921 /*
1922  * Check whether we are dealing with Arabic combining characters.
1923  * Note: these are NOT really composing characters!
1924  */
1925     int
1926 arabic_combine(
1927     int		one,	    /* first character */
1928     int		two)	    /* character just after "one" */
1929 {
1930     if (one == a_LAM)
1931 	return arabic_maycombine(two);
1932     return FALSE;
1933 }
1934 
1935 /*
1936  * Check whether we are dealing with a character that could be regarded as an
1937  * Arabic combining character, need to check the character before this.
1938  */
1939     int
1940 arabic_maycombine(int two)
1941 {
1942     if (p_arshape && !p_tbidi)
1943 	return (two == a_ALEF_MADDA
1944 		    || two == a_ALEF_HAMZA_ABOVE
1945 		    || two == a_ALEF_HAMZA_BELOW
1946 		    || two == a_ALEF);
1947     return FALSE;
1948 }
1949 
1950 /*
1951  * Check if the character pointed to by "p2" is a composing character when it
1952  * comes after "p1".  For Arabic sometimes "ab" is replaced with "c", which
1953  * behaves like a composing character.
1954  */
1955     int
1956 utf_composinglike(char_u *p1, char_u *p2)
1957 {
1958     int		c2;
1959 
1960     c2 = utf_ptr2char(p2);
1961     if (utf_iscomposing(c2))
1962 	return TRUE;
1963     if (!arabic_maycombine(c2))
1964 	return FALSE;
1965     return arabic_combine(utf_ptr2char(p1), c2);
1966 }
1967 #endif
1968 
1969 /*
1970  * Convert a UTF-8 byte string to a wide character.  Also get up to MAX_MCO
1971  * composing characters.
1972  */
1973     int
1974 utfc_ptr2char(
1975     char_u	*p,
1976     int		*pcc)	/* return: composing chars, last one is 0 */
1977 {
1978     int		len;
1979     int		c;
1980     int		cc;
1981     int		i = 0;
1982 
1983     c = utf_ptr2char(p);
1984     len = utf_ptr2len(p);
1985 
1986     /* Only accept a composing char when the first char isn't illegal. */
1987     if ((len > 1 || *p < 0x80)
1988 	    && p[len] >= 0x80
1989 	    && UTF_COMPOSINGLIKE(p, p + len))
1990     {
1991 	cc = utf_ptr2char(p + len);
1992 	for (;;)
1993 	{
1994 	    pcc[i++] = cc;
1995 	    if (i == MAX_MCO)
1996 		break;
1997 	    len += utf_ptr2len(p + len);
1998 	    if (p[len] < 0x80 || !utf_iscomposing(cc = utf_ptr2char(p + len)))
1999 		break;
2000 	}
2001     }
2002 
2003     if (i < MAX_MCO)	/* last composing char must be 0 */
2004 	pcc[i] = 0;
2005 
2006     return c;
2007 }
2008 
2009 /*
2010  * Convert a UTF-8 byte string to a wide character.  Also get up to MAX_MCO
2011  * composing characters.  Use no more than p[maxlen].
2012  */
2013     int
2014 utfc_ptr2char_len(
2015     char_u	*p,
2016     int		*pcc,	/* return: composing chars, last one is 0 */
2017     int		maxlen)
2018 {
2019     int		len;
2020     int		c;
2021     int		cc;
2022     int		i = 0;
2023 
2024     c = utf_ptr2char(p);
2025     len = utf_ptr2len_len(p, maxlen);
2026     /* Only accept a composing char when the first char isn't illegal. */
2027     if ((len > 1 || *p < 0x80)
2028 	    && len < maxlen
2029 	    && p[len] >= 0x80
2030 	    && UTF_COMPOSINGLIKE(p, p + len))
2031     {
2032 	cc = utf_ptr2char(p + len);
2033 	for (;;)
2034 	{
2035 	    pcc[i++] = cc;
2036 	    if (i == MAX_MCO)
2037 		break;
2038 	    len += utf_ptr2len_len(p + len, maxlen - len);
2039 	    if (len >= maxlen
2040 		    || p[len] < 0x80
2041 		    || !utf_iscomposing(cc = utf_ptr2char(p + len)))
2042 		break;
2043 	}
2044     }
2045 
2046     if (i < MAX_MCO)	/* last composing char must be 0 */
2047 	pcc[i] = 0;
2048 
2049     return c;
2050 }
2051 
2052 /*
2053  * Convert the character at screen position "off" to a sequence of bytes.
2054  * Includes the composing characters.
2055  * "buf" must at least have the length MB_MAXBYTES + 1.
2056  * Only to be used when ScreenLinesUC[off] != 0.
2057  * Returns the produced number of bytes.
2058  */
2059     int
2060 utfc_char2bytes(int off, char_u *buf)
2061 {
2062     int		len;
2063     int		i;
2064 
2065     len = utf_char2bytes(ScreenLinesUC[off], buf);
2066     for (i = 0; i < Screen_mco; ++i)
2067     {
2068 	if (ScreenLinesC[i][off] == 0)
2069 	    break;
2070 	len += utf_char2bytes(ScreenLinesC[i][off], buf + len);
2071     }
2072     return len;
2073 }
2074 
2075 /*
2076  * Get the length of a UTF-8 byte sequence, not including any following
2077  * composing characters.
2078  * Returns 0 for "".
2079  * Returns 1 for an illegal byte sequence.
2080  */
2081     int
2082 utf_ptr2len(char_u *p)
2083 {
2084     int		len;
2085     int		i;
2086 
2087     if (*p == NUL)
2088 	return 0;
2089     len = utf8len_tab[*p];
2090     for (i = 1; i < len; ++i)
2091 	if ((p[i] & 0xc0) != 0x80)
2092 	    return 1;
2093     return len;
2094 }
2095 
2096 /*
2097  * Return length of UTF-8 character, obtained from the first byte.
2098  * "b" must be between 0 and 255!
2099  * Returns 1 for an invalid first byte value.
2100  */
2101     int
2102 utf_byte2len(int b)
2103 {
2104     return utf8len_tab[b];
2105 }
2106 
2107 /*
2108  * Get the length of UTF-8 byte sequence "p[size]".  Does not include any
2109  * following composing characters.
2110  * Returns 1 for "".
2111  * Returns 1 for an illegal byte sequence (also in incomplete byte seq.).
2112  * Returns number > "size" for an incomplete byte sequence.
2113  * Never returns zero.
2114  */
2115     int
2116 utf_ptr2len_len(char_u *p, int size)
2117 {
2118     int		len;
2119     int		i;
2120     int		m;
2121 
2122     len = utf8len_tab[*p];
2123     if (len == 1)
2124 	return 1;	/* NUL, ascii or illegal lead byte */
2125     if (len > size)
2126 	m = size;	/* incomplete byte sequence. */
2127     else
2128 	m = len;
2129     for (i = 1; i < m; ++i)
2130 	if ((p[i] & 0xc0) != 0x80)
2131 	    return 1;
2132     return len;
2133 }
2134 
2135 /*
2136  * Return the number of bytes the UTF-8 encoding of the character at "p" takes.
2137  * This includes following composing characters.
2138  */
2139     int
2140 utfc_ptr2len(char_u *p)
2141 {
2142     int		len;
2143     int		b0 = *p;
2144 #ifdef FEAT_ARABIC
2145     int		prevlen;
2146 #endif
2147 
2148     if (b0 == NUL)
2149 	return 0;
2150     if (b0 < 0x80 && p[1] < 0x80)	/* be quick for ASCII */
2151 	return 1;
2152 
2153     /* Skip over first UTF-8 char, stopping at a NUL byte. */
2154     len = utf_ptr2len(p);
2155 
2156     /* Check for illegal byte. */
2157     if (len == 1 && b0 >= 0x80)
2158 	return 1;
2159 
2160     /*
2161      * Check for composing characters.  We can handle only the first six, but
2162      * skip all of them (otherwise the cursor would get stuck).
2163      */
2164 #ifdef FEAT_ARABIC
2165     prevlen = 0;
2166 #endif
2167     for (;;)
2168     {
2169 	if (p[len] < 0x80 || !UTF_COMPOSINGLIKE(p + prevlen, p + len))
2170 	    return len;
2171 
2172 	/* Skip over composing char */
2173 #ifdef FEAT_ARABIC
2174 	prevlen = len;
2175 #endif
2176 	len += utf_ptr2len(p + len);
2177     }
2178 }
2179 
2180 /*
2181  * Return the number of bytes the UTF-8 encoding of the character at "p[size]"
2182  * takes.  This includes following composing characters.
2183  * Returns 0 for an empty string.
2184  * Returns 1 for an illegal char or an incomplete byte sequence.
2185  */
2186     int
2187 utfc_ptr2len_len(char_u *p, int size)
2188 {
2189     int		len;
2190 #ifdef FEAT_ARABIC
2191     int		prevlen;
2192 #endif
2193 
2194     if (size < 1 || *p == NUL)
2195 	return 0;
2196     if (p[0] < 0x80 && (size == 1 || p[1] < 0x80)) /* be quick for ASCII */
2197 	return 1;
2198 
2199     /* Skip over first UTF-8 char, stopping at a NUL byte. */
2200     len = utf_ptr2len_len(p, size);
2201 
2202     /* Check for illegal byte and incomplete byte sequence. */
2203     if ((len == 1 && p[0] >= 0x80) || len > size)
2204 	return 1;
2205 
2206     /*
2207      * Check for composing characters.  We can handle only the first six, but
2208      * skip all of them (otherwise the cursor would get stuck).
2209      */
2210 #ifdef FEAT_ARABIC
2211     prevlen = 0;
2212 #endif
2213     while (len < size)
2214     {
2215 	int	len_next_char;
2216 
2217 	if (p[len] < 0x80)
2218 	    break;
2219 
2220 	/*
2221 	 * Next character length should not go beyond size to ensure that
2222 	 * UTF_COMPOSINGLIKE(...) does not read beyond size.
2223 	 */
2224 	len_next_char = utf_ptr2len_len(p + len, size - len);
2225 	if (len_next_char > size - len)
2226 	    break;
2227 
2228 	if (!UTF_COMPOSINGLIKE(p + prevlen, p + len))
2229 	    break;
2230 
2231 	/* Skip over composing char */
2232 #ifdef FEAT_ARABIC
2233 	prevlen = len;
2234 #endif
2235 	len += len_next_char;
2236     }
2237     return len;
2238 }
2239 
2240 /*
2241  * Return the number of bytes the UTF-8 encoding of character "c" takes.
2242  * This does not include composing characters.
2243  */
2244     int
2245 utf_char2len(int c)
2246 {
2247     if (c < 0x80)
2248 	return 1;
2249     if (c < 0x800)
2250 	return 2;
2251     if (c < 0x10000)
2252 	return 3;
2253     if (c < 0x200000)
2254 	return 4;
2255     if (c < 0x4000000)
2256 	return 5;
2257     return 6;
2258 }
2259 
2260 /*
2261  * Convert Unicode character "c" to UTF-8 string in "buf[]".
2262  * Returns the number of bytes.
2263  */
2264     int
2265 utf_char2bytes(int c, char_u *buf)
2266 {
2267     if (c < 0x80)		/* 7 bits */
2268     {
2269 	buf[0] = c;
2270 	return 1;
2271     }
2272     if (c < 0x800)		/* 11 bits */
2273     {
2274 	buf[0] = 0xc0 + ((unsigned)c >> 6);
2275 	buf[1] = 0x80 + (c & 0x3f);
2276 	return 2;
2277     }
2278     if (c < 0x10000)		/* 16 bits */
2279     {
2280 	buf[0] = 0xe0 + ((unsigned)c >> 12);
2281 	buf[1] = 0x80 + (((unsigned)c >> 6) & 0x3f);
2282 	buf[2] = 0x80 + (c & 0x3f);
2283 	return 3;
2284     }
2285     if (c < 0x200000)		/* 21 bits */
2286     {
2287 	buf[0] = 0xf0 + ((unsigned)c >> 18);
2288 	buf[1] = 0x80 + (((unsigned)c >> 12) & 0x3f);
2289 	buf[2] = 0x80 + (((unsigned)c >> 6) & 0x3f);
2290 	buf[3] = 0x80 + (c & 0x3f);
2291 	return 4;
2292     }
2293     if (c < 0x4000000)		/* 26 bits */
2294     {
2295 	buf[0] = 0xf8 + ((unsigned)c >> 24);
2296 	buf[1] = 0x80 + (((unsigned)c >> 18) & 0x3f);
2297 	buf[2] = 0x80 + (((unsigned)c >> 12) & 0x3f);
2298 	buf[3] = 0x80 + (((unsigned)c >> 6) & 0x3f);
2299 	buf[4] = 0x80 + (c & 0x3f);
2300 	return 5;
2301     }
2302 				/* 31 bits */
2303     buf[0] = 0xfc + ((unsigned)c >> 30);
2304     buf[1] = 0x80 + (((unsigned)c >> 24) & 0x3f);
2305     buf[2] = 0x80 + (((unsigned)c >> 18) & 0x3f);
2306     buf[3] = 0x80 + (((unsigned)c >> 12) & 0x3f);
2307     buf[4] = 0x80 + (((unsigned)c >> 6) & 0x3f);
2308     buf[5] = 0x80 + (c & 0x3f);
2309     return 6;
2310 }
2311 
2312 #if defined(FEAT_TERMINAL) || defined(PROTO)
2313 /*
2314  * utf_iscomposing() with different argument type for libvterm.
2315  */
2316     int
2317 utf_iscomposing_uint(UINT32_T c)
2318 {
2319     return utf_iscomposing((int)c);
2320 }
2321 #endif
2322 
2323 /*
2324  * Return TRUE if "c" is a composing UTF-8 character.  This means it will be
2325  * drawn on top of the preceding character.
2326  * Based on code from Markus Kuhn.
2327  */
2328     int
2329 utf_iscomposing(int c)
2330 {
2331     /* Sorted list of non-overlapping intervals.
2332      * Generated by ../runtime/tools/unicode.vim. */
2333     static struct interval combining[] =
2334     {
2335 	{0x0300, 0x036f},
2336 	{0x0483, 0x0489},
2337 	{0x0591, 0x05bd},
2338 	{0x05bf, 0x05bf},
2339 	{0x05c1, 0x05c2},
2340 	{0x05c4, 0x05c5},
2341 	{0x05c7, 0x05c7},
2342 	{0x0610, 0x061a},
2343 	{0x064b, 0x065f},
2344 	{0x0670, 0x0670},
2345 	{0x06d6, 0x06dc},
2346 	{0x06df, 0x06e4},
2347 	{0x06e7, 0x06e8},
2348 	{0x06ea, 0x06ed},
2349 	{0x0711, 0x0711},
2350 	{0x0730, 0x074a},
2351 	{0x07a6, 0x07b0},
2352 	{0x07eb, 0x07f3},
2353 	{0x07fd, 0x07fd},
2354 	{0x0816, 0x0819},
2355 	{0x081b, 0x0823},
2356 	{0x0825, 0x0827},
2357 	{0x0829, 0x082d},
2358 	{0x0859, 0x085b},
2359 	{0x08d3, 0x08e1},
2360 	{0x08e3, 0x0903},
2361 	{0x093a, 0x093c},
2362 	{0x093e, 0x094f},
2363 	{0x0951, 0x0957},
2364 	{0x0962, 0x0963},
2365 	{0x0981, 0x0983},
2366 	{0x09bc, 0x09bc},
2367 	{0x09be, 0x09c4},
2368 	{0x09c7, 0x09c8},
2369 	{0x09cb, 0x09cd},
2370 	{0x09d7, 0x09d7},
2371 	{0x09e2, 0x09e3},
2372 	{0x09fe, 0x09fe},
2373 	{0x0a01, 0x0a03},
2374 	{0x0a3c, 0x0a3c},
2375 	{0x0a3e, 0x0a42},
2376 	{0x0a47, 0x0a48},
2377 	{0x0a4b, 0x0a4d},
2378 	{0x0a51, 0x0a51},
2379 	{0x0a70, 0x0a71},
2380 	{0x0a75, 0x0a75},
2381 	{0x0a81, 0x0a83},
2382 	{0x0abc, 0x0abc},
2383 	{0x0abe, 0x0ac5},
2384 	{0x0ac7, 0x0ac9},
2385 	{0x0acb, 0x0acd},
2386 	{0x0ae2, 0x0ae3},
2387 	{0x0afa, 0x0aff},
2388 	{0x0b01, 0x0b03},
2389 	{0x0b3c, 0x0b3c},
2390 	{0x0b3e, 0x0b44},
2391 	{0x0b47, 0x0b48},
2392 	{0x0b4b, 0x0b4d},
2393 	{0x0b56, 0x0b57},
2394 	{0x0b62, 0x0b63},
2395 	{0x0b82, 0x0b82},
2396 	{0x0bbe, 0x0bc2},
2397 	{0x0bc6, 0x0bc8},
2398 	{0x0bca, 0x0bcd},
2399 	{0x0bd7, 0x0bd7},
2400 	{0x0c00, 0x0c04},
2401 	{0x0c3e, 0x0c44},
2402 	{0x0c46, 0x0c48},
2403 	{0x0c4a, 0x0c4d},
2404 	{0x0c55, 0x0c56},
2405 	{0x0c62, 0x0c63},
2406 	{0x0c81, 0x0c83},
2407 	{0x0cbc, 0x0cbc},
2408 	{0x0cbe, 0x0cc4},
2409 	{0x0cc6, 0x0cc8},
2410 	{0x0cca, 0x0ccd},
2411 	{0x0cd5, 0x0cd6},
2412 	{0x0ce2, 0x0ce3},
2413 	{0x0d00, 0x0d03},
2414 	{0x0d3b, 0x0d3c},
2415 	{0x0d3e, 0x0d44},
2416 	{0x0d46, 0x0d48},
2417 	{0x0d4a, 0x0d4d},
2418 	{0x0d57, 0x0d57},
2419 	{0x0d62, 0x0d63},
2420 	{0x0d82, 0x0d83},
2421 	{0x0dca, 0x0dca},
2422 	{0x0dcf, 0x0dd4},
2423 	{0x0dd6, 0x0dd6},
2424 	{0x0dd8, 0x0ddf},
2425 	{0x0df2, 0x0df3},
2426 	{0x0e31, 0x0e31},
2427 	{0x0e34, 0x0e3a},
2428 	{0x0e47, 0x0e4e},
2429 	{0x0eb1, 0x0eb1},
2430 	{0x0eb4, 0x0eb9},
2431 	{0x0ebb, 0x0ebc},
2432 	{0x0ec8, 0x0ecd},
2433 	{0x0f18, 0x0f19},
2434 	{0x0f35, 0x0f35},
2435 	{0x0f37, 0x0f37},
2436 	{0x0f39, 0x0f39},
2437 	{0x0f3e, 0x0f3f},
2438 	{0x0f71, 0x0f84},
2439 	{0x0f86, 0x0f87},
2440 	{0x0f8d, 0x0f97},
2441 	{0x0f99, 0x0fbc},
2442 	{0x0fc6, 0x0fc6},
2443 	{0x102b, 0x103e},
2444 	{0x1056, 0x1059},
2445 	{0x105e, 0x1060},
2446 	{0x1062, 0x1064},
2447 	{0x1067, 0x106d},
2448 	{0x1071, 0x1074},
2449 	{0x1082, 0x108d},
2450 	{0x108f, 0x108f},
2451 	{0x109a, 0x109d},
2452 	{0x135d, 0x135f},
2453 	{0x1712, 0x1714},
2454 	{0x1732, 0x1734},
2455 	{0x1752, 0x1753},
2456 	{0x1772, 0x1773},
2457 	{0x17b4, 0x17d3},
2458 	{0x17dd, 0x17dd},
2459 	{0x180b, 0x180d},
2460 	{0x1885, 0x1886},
2461 	{0x18a9, 0x18a9},
2462 	{0x1920, 0x192b},
2463 	{0x1930, 0x193b},
2464 	{0x1a17, 0x1a1b},
2465 	{0x1a55, 0x1a5e},
2466 	{0x1a60, 0x1a7c},
2467 	{0x1a7f, 0x1a7f},
2468 	{0x1ab0, 0x1abe},
2469 	{0x1b00, 0x1b04},
2470 	{0x1b34, 0x1b44},
2471 	{0x1b6b, 0x1b73},
2472 	{0x1b80, 0x1b82},
2473 	{0x1ba1, 0x1bad},
2474 	{0x1be6, 0x1bf3},
2475 	{0x1c24, 0x1c37},
2476 	{0x1cd0, 0x1cd2},
2477 	{0x1cd4, 0x1ce8},
2478 	{0x1ced, 0x1ced},
2479 	{0x1cf2, 0x1cf4},
2480 	{0x1cf7, 0x1cf9},
2481 	{0x1dc0, 0x1df9},
2482 	{0x1dfb, 0x1dff},
2483 	{0x20d0, 0x20f0},
2484 	{0x2cef, 0x2cf1},
2485 	{0x2d7f, 0x2d7f},
2486 	{0x2de0, 0x2dff},
2487 	{0x302a, 0x302f},
2488 	{0x3099, 0x309a},
2489 	{0xa66f, 0xa672},
2490 	{0xa674, 0xa67d},
2491 	{0xa69e, 0xa69f},
2492 	{0xa6f0, 0xa6f1},
2493 	{0xa802, 0xa802},
2494 	{0xa806, 0xa806},
2495 	{0xa80b, 0xa80b},
2496 	{0xa823, 0xa827},
2497 	{0xa880, 0xa881},
2498 	{0xa8b4, 0xa8c5},
2499 	{0xa8e0, 0xa8f1},
2500 	{0xa8ff, 0xa8ff},
2501 	{0xa926, 0xa92d},
2502 	{0xa947, 0xa953},
2503 	{0xa980, 0xa983},
2504 	{0xa9b3, 0xa9c0},
2505 	{0xa9e5, 0xa9e5},
2506 	{0xaa29, 0xaa36},
2507 	{0xaa43, 0xaa43},
2508 	{0xaa4c, 0xaa4d},
2509 	{0xaa7b, 0xaa7d},
2510 	{0xaab0, 0xaab0},
2511 	{0xaab2, 0xaab4},
2512 	{0xaab7, 0xaab8},
2513 	{0xaabe, 0xaabf},
2514 	{0xaac1, 0xaac1},
2515 	{0xaaeb, 0xaaef},
2516 	{0xaaf5, 0xaaf6},
2517 	{0xabe3, 0xabea},
2518 	{0xabec, 0xabed},
2519 	{0xfb1e, 0xfb1e},
2520 	{0xfe00, 0xfe0f},
2521 	{0xfe20, 0xfe2f},
2522 	{0x101fd, 0x101fd},
2523 	{0x102e0, 0x102e0},
2524 	{0x10376, 0x1037a},
2525 	{0x10a01, 0x10a03},
2526 	{0x10a05, 0x10a06},
2527 	{0x10a0c, 0x10a0f},
2528 	{0x10a38, 0x10a3a},
2529 	{0x10a3f, 0x10a3f},
2530 	{0x10ae5, 0x10ae6},
2531 	{0x10d24, 0x10d27},
2532 	{0x10f46, 0x10f50},
2533 	{0x11000, 0x11002},
2534 	{0x11038, 0x11046},
2535 	{0x1107f, 0x11082},
2536 	{0x110b0, 0x110ba},
2537 	{0x11100, 0x11102},
2538 	{0x11127, 0x11134},
2539 	{0x11145, 0x11146},
2540 	{0x11173, 0x11173},
2541 	{0x11180, 0x11182},
2542 	{0x111b3, 0x111c0},
2543 	{0x111c9, 0x111cc},
2544 	{0x1122c, 0x11237},
2545 	{0x1123e, 0x1123e},
2546 	{0x112df, 0x112ea},
2547 	{0x11300, 0x11303},
2548 	{0x1133b, 0x1133c},
2549 	{0x1133e, 0x11344},
2550 	{0x11347, 0x11348},
2551 	{0x1134b, 0x1134d},
2552 	{0x11357, 0x11357},
2553 	{0x11362, 0x11363},
2554 	{0x11366, 0x1136c},
2555 	{0x11370, 0x11374},
2556 	{0x11435, 0x11446},
2557 	{0x1145e, 0x1145e},
2558 	{0x114b0, 0x114c3},
2559 	{0x115af, 0x115b5},
2560 	{0x115b8, 0x115c0},
2561 	{0x115dc, 0x115dd},
2562 	{0x11630, 0x11640},
2563 	{0x116ab, 0x116b7},
2564 	{0x1171d, 0x1172b},
2565 	{0x1182c, 0x1183a},
2566 	{0x11a01, 0x11a0a},
2567 	{0x11a33, 0x11a39},
2568 	{0x11a3b, 0x11a3e},
2569 	{0x11a47, 0x11a47},
2570 	{0x11a51, 0x11a5b},
2571 	{0x11a8a, 0x11a99},
2572 	{0x11c2f, 0x11c36},
2573 	{0x11c38, 0x11c3f},
2574 	{0x11c92, 0x11ca7},
2575 	{0x11ca9, 0x11cb6},
2576 	{0x11d31, 0x11d36},
2577 	{0x11d3a, 0x11d3a},
2578 	{0x11d3c, 0x11d3d},
2579 	{0x11d3f, 0x11d45},
2580 	{0x11d47, 0x11d47},
2581 	{0x11d8a, 0x11d8e},
2582 	{0x11d90, 0x11d91},
2583 	{0x11d93, 0x11d97},
2584 	{0x11ef3, 0x11ef6},
2585 	{0x16af0, 0x16af4},
2586 	{0x16b30, 0x16b36},
2587 	{0x16f51, 0x16f7e},
2588 	{0x16f8f, 0x16f92},
2589 	{0x1bc9d, 0x1bc9e},
2590 	{0x1d165, 0x1d169},
2591 	{0x1d16d, 0x1d172},
2592 	{0x1d17b, 0x1d182},
2593 	{0x1d185, 0x1d18b},
2594 	{0x1d1aa, 0x1d1ad},
2595 	{0x1d242, 0x1d244},
2596 	{0x1da00, 0x1da36},
2597 	{0x1da3b, 0x1da6c},
2598 	{0x1da75, 0x1da75},
2599 	{0x1da84, 0x1da84},
2600 	{0x1da9b, 0x1da9f},
2601 	{0x1daa1, 0x1daaf},
2602 	{0x1e000, 0x1e006},
2603 	{0x1e008, 0x1e018},
2604 	{0x1e01b, 0x1e021},
2605 	{0x1e023, 0x1e024},
2606 	{0x1e026, 0x1e02a},
2607 	{0x1e8d0, 0x1e8d6},
2608 	{0x1e944, 0x1e94a},
2609 	{0xe0100, 0xe01ef}
2610     };
2611 
2612     return intable(combining, sizeof(combining), c);
2613 }
2614 
2615 /*
2616  * Return TRUE for characters that can be displayed in a normal way.
2617  * Only for characters of 0x100 and above!
2618  */
2619     int
2620 utf_printable(int c)
2621 {
2622 #ifdef USE_WCHAR_FUNCTIONS
2623     /*
2624      * Assume the iswprint() library function works better than our own stuff.
2625      */
2626     return iswprint(c);
2627 #else
2628     /* Sorted list of non-overlapping intervals.
2629      * 0xd800-0xdfff is reserved for UTF-16, actually illegal. */
2630     static struct interval nonprint[] =
2631     {
2632 	{0x070f, 0x070f}, {0x180b, 0x180e}, {0x200b, 0x200f}, {0x202a, 0x202e},
2633 	{0x206a, 0x206f}, {0xd800, 0xdfff}, {0xfeff, 0xfeff}, {0xfff9, 0xfffb},
2634 	{0xfffe, 0xffff}
2635     };
2636 
2637     return !intable(nonprint, sizeof(nonprint), c);
2638 #endif
2639 }
2640 
2641 /* Sorted list of non-overlapping intervals of all Emoji characters,
2642  * based on http://unicode.org/emoji/charts/emoji-list.html */
2643 static struct interval emoji_all[] =
2644 {
2645     {0x203c, 0x203c},
2646     {0x2049, 0x2049},
2647     {0x2122, 0x2122},
2648     {0x2139, 0x2139},
2649     {0x2194, 0x2199},
2650     {0x21a9, 0x21aa},
2651     {0x231a, 0x231b},
2652     {0x2328, 0x2328},
2653     {0x23cf, 0x23cf},
2654     {0x23e9, 0x23f3},
2655     {0x23f8, 0x23fa},
2656     {0x24c2, 0x24c2},
2657     {0x25aa, 0x25ab},
2658     {0x25b6, 0x25b6},
2659     {0x25c0, 0x25c0},
2660     {0x25fb, 0x25fe},
2661     {0x2600, 0x2604},
2662     {0x260e, 0x260e},
2663     {0x2611, 0x2611},
2664     {0x2614, 0x2615},
2665     {0x2618, 0x2618},
2666     {0x261d, 0x261d},
2667     {0x2620, 0x2620},
2668     {0x2622, 0x2623},
2669     {0x2626, 0x2626},
2670     {0x262a, 0x262a},
2671     {0x262e, 0x262f},
2672     {0x2638, 0x263a},
2673     {0x2640, 0x2640},
2674     {0x2642, 0x2642},
2675     {0x2648, 0x2653},
2676     {0x265f, 0x2660},
2677     {0x2663, 0x2663},
2678     {0x2665, 0x2666},
2679     {0x2668, 0x2668},
2680     {0x267b, 0x267b},
2681     {0x267e, 0x267f},
2682     {0x2692, 0x2697},
2683     {0x2699, 0x2699},
2684     {0x269b, 0x269c},
2685     {0x26a0, 0x26a1},
2686     {0x26aa, 0x26ab},
2687     {0x26b0, 0x26b1},
2688     {0x26bd, 0x26be},
2689     {0x26c4, 0x26c5},
2690     {0x26c8, 0x26c8},
2691     {0x26ce, 0x26cf},
2692     {0x26d1, 0x26d1},
2693     {0x26d3, 0x26d4},
2694     {0x26e9, 0x26ea},
2695     {0x26f0, 0x26f5},
2696     {0x26f7, 0x26fa},
2697     {0x26fd, 0x26fd},
2698     {0x2702, 0x2702},
2699     {0x2705, 0x2705},
2700     {0x2708, 0x270d},
2701     {0x270f, 0x270f},
2702     {0x2712, 0x2712},
2703     {0x2714, 0x2714},
2704     {0x2716, 0x2716},
2705     {0x271d, 0x271d},
2706     {0x2721, 0x2721},
2707     {0x2728, 0x2728},
2708     {0x2733, 0x2734},
2709     {0x2744, 0x2744},
2710     {0x2747, 0x2747},
2711     {0x274c, 0x274c},
2712     {0x274e, 0x274e},
2713     {0x2753, 0x2755},
2714     {0x2757, 0x2757},
2715     {0x2763, 0x2764},
2716     {0x2795, 0x2797},
2717     {0x27a1, 0x27a1},
2718     {0x27b0, 0x27b0},
2719     {0x27bf, 0x27bf},
2720     {0x2934, 0x2935},
2721     {0x2b05, 0x2b07},
2722     {0x2b1b, 0x2b1c},
2723     {0x2b50, 0x2b50},
2724     {0x2b55, 0x2b55},
2725     {0x3030, 0x3030},
2726     {0x303d, 0x303d},
2727     {0x3297, 0x3297},
2728     {0x3299, 0x3299},
2729     {0x1f004, 0x1f004},
2730     {0x1f0cf, 0x1f0cf},
2731     {0x1f170, 0x1f171},
2732     {0x1f17e, 0x1f17f},
2733     {0x1f18e, 0x1f18e},
2734     {0x1f191, 0x1f19a},
2735     {0x1f1e6, 0x1f1ff},
2736     {0x1f201, 0x1f202},
2737     {0x1f21a, 0x1f21a},
2738     {0x1f22f, 0x1f22f},
2739     {0x1f232, 0x1f23a},
2740     {0x1f250, 0x1f251},
2741     {0x1f300, 0x1f321},
2742     {0x1f324, 0x1f393},
2743     {0x1f396, 0x1f397},
2744     {0x1f399, 0x1f39b},
2745     {0x1f39e, 0x1f3f0},
2746     {0x1f3f3, 0x1f3f5},
2747     {0x1f3f7, 0x1f4fd},
2748     {0x1f4ff, 0x1f53d},
2749     {0x1f549, 0x1f54e},
2750     {0x1f550, 0x1f567},
2751     {0x1f56f, 0x1f570},
2752     {0x1f573, 0x1f57a},
2753     {0x1f587, 0x1f587},
2754     {0x1f58a, 0x1f58d},
2755     {0x1f590, 0x1f590},
2756     {0x1f595, 0x1f596},
2757     {0x1f5a4, 0x1f5a5},
2758     {0x1f5a8, 0x1f5a8},
2759     {0x1f5b1, 0x1f5b2},
2760     {0x1f5bc, 0x1f5bc},
2761     {0x1f5c2, 0x1f5c4},
2762     {0x1f5d1, 0x1f5d3},
2763     {0x1f5dc, 0x1f5de},
2764     {0x1f5e1, 0x1f5e1},
2765     {0x1f5e3, 0x1f5e3},
2766     {0x1f5e8, 0x1f5e8},
2767     {0x1f5ef, 0x1f5ef},
2768     {0x1f5f3, 0x1f5f3},
2769     {0x1f5fa, 0x1f64f},
2770     {0x1f680, 0x1f6c5},
2771     {0x1f6cb, 0x1f6d2},
2772     {0x1f6e0, 0x1f6e5},
2773     {0x1f6e9, 0x1f6e9},
2774     {0x1f6eb, 0x1f6ec},
2775     {0x1f6f0, 0x1f6f0},
2776     {0x1f6f3, 0x1f6f9},
2777     {0x1f910, 0x1f93a},
2778     {0x1f93c, 0x1f93e},
2779     {0x1f940, 0x1f945},
2780     {0x1f947, 0x1f970},
2781     {0x1f973, 0x1f976},
2782     {0x1f97a, 0x1f97a},
2783     {0x1f97c, 0x1f9a2},
2784     {0x1f9b0, 0x1f9b9},
2785     {0x1f9c0, 0x1f9c2},
2786     {0x1f9d0, 0x1f9ff}
2787 };
2788 
2789 /*
2790  * Get class of a Unicode character.
2791  * 0: white space
2792  * 1: punctuation
2793  * 2 or bigger: some class of word character.
2794  */
2795     int
2796 utf_class(int c)
2797 {
2798     return utf_class_buf(c, curbuf);
2799 }
2800 
2801     int
2802 utf_class_buf(int c, buf_T *buf)
2803 {
2804     /* sorted list of non-overlapping intervals */
2805     static struct clinterval
2806     {
2807 	unsigned int first;
2808 	unsigned int last;
2809 	unsigned int class;
2810     } classes[] =
2811     {
2812 	{0x037e, 0x037e, 1},		/* Greek question mark */
2813 	{0x0387, 0x0387, 1},		/* Greek ano teleia */
2814 	{0x055a, 0x055f, 1},		/* Armenian punctuation */
2815 	{0x0589, 0x0589, 1},		/* Armenian full stop */
2816 	{0x05be, 0x05be, 1},
2817 	{0x05c0, 0x05c0, 1},
2818 	{0x05c3, 0x05c3, 1},
2819 	{0x05f3, 0x05f4, 1},
2820 	{0x060c, 0x060c, 1},
2821 	{0x061b, 0x061b, 1},
2822 	{0x061f, 0x061f, 1},
2823 	{0x066a, 0x066d, 1},
2824 	{0x06d4, 0x06d4, 1},
2825 	{0x0700, 0x070d, 1},		/* Syriac punctuation */
2826 	{0x0964, 0x0965, 1},
2827 	{0x0970, 0x0970, 1},
2828 	{0x0df4, 0x0df4, 1},
2829 	{0x0e4f, 0x0e4f, 1},
2830 	{0x0e5a, 0x0e5b, 1},
2831 	{0x0f04, 0x0f12, 1},
2832 	{0x0f3a, 0x0f3d, 1},
2833 	{0x0f85, 0x0f85, 1},
2834 	{0x104a, 0x104f, 1},		/* Myanmar punctuation */
2835 	{0x10fb, 0x10fb, 1},		/* Georgian punctuation */
2836 	{0x1361, 0x1368, 1},		/* Ethiopic punctuation */
2837 	{0x166d, 0x166e, 1},		/* Canadian Syl. punctuation */
2838 	{0x1680, 0x1680, 0},
2839 	{0x169b, 0x169c, 1},
2840 	{0x16eb, 0x16ed, 1},
2841 	{0x1735, 0x1736, 1},
2842 	{0x17d4, 0x17dc, 1},		/* Khmer punctuation */
2843 	{0x1800, 0x180a, 1},		/* Mongolian punctuation */
2844 	{0x2000, 0x200b, 0},		/* spaces */
2845 	{0x200c, 0x2027, 1},		/* punctuation and symbols */
2846 	{0x2028, 0x2029, 0},
2847 	{0x202a, 0x202e, 1},		/* punctuation and symbols */
2848 	{0x202f, 0x202f, 0},
2849 	{0x2030, 0x205e, 1},		/* punctuation and symbols */
2850 	{0x205f, 0x205f, 0},
2851 	{0x2060, 0x27ff, 1},		/* punctuation and symbols */
2852 	{0x2070, 0x207f, 0x2070},	/* superscript */
2853 	{0x2080, 0x2094, 0x2080},	/* subscript */
2854 	{0x20a0, 0x27ff, 1},		/* all kinds of symbols */
2855 	{0x2800, 0x28ff, 0x2800},	/* braille */
2856 	{0x2900, 0x2998, 1},		/* arrows, brackets, etc. */
2857 	{0x29d8, 0x29db, 1},
2858 	{0x29fc, 0x29fd, 1},
2859 	{0x2e00, 0x2e7f, 1},		/* supplemental punctuation */
2860 	{0x3000, 0x3000, 0},		/* ideographic space */
2861 	{0x3001, 0x3020, 1},		/* ideographic punctuation */
2862 	{0x3030, 0x3030, 1},
2863 	{0x303d, 0x303d, 1},
2864 	{0x3040, 0x309f, 0x3040},	/* Hiragana */
2865 	{0x30a0, 0x30ff, 0x30a0},	/* Katakana */
2866 	{0x3300, 0x9fff, 0x4e00},	/* CJK Ideographs */
2867 	{0xac00, 0xd7a3, 0xac00},	/* Hangul Syllables */
2868 	{0xf900, 0xfaff, 0x4e00},	/* CJK Ideographs */
2869 	{0xfd3e, 0xfd3f, 1},
2870 	{0xfe30, 0xfe6b, 1},		/* punctuation forms */
2871 	{0xff00, 0xff0f, 1},		/* half/fullwidth ASCII */
2872 	{0xff1a, 0xff20, 1},		/* half/fullwidth ASCII */
2873 	{0xff3b, 0xff40, 1},		/* half/fullwidth ASCII */
2874 	{0xff5b, 0xff65, 1},		/* half/fullwidth ASCII */
2875 	{0x20000, 0x2a6df, 0x4e00},	/* CJK Ideographs */
2876 	{0x2a700, 0x2b73f, 0x4e00},	/* CJK Ideographs */
2877 	{0x2b740, 0x2b81f, 0x4e00},	/* CJK Ideographs */
2878 	{0x2f800, 0x2fa1f, 0x4e00},	/* CJK Ideographs */
2879     };
2880 
2881     int bot = 0;
2882     int top = sizeof(classes) / sizeof(struct clinterval) - 1;
2883     int mid;
2884 
2885     /* First quick check for Latin1 characters, use 'iskeyword'. */
2886     if (c < 0x100)
2887     {
2888 	if (c == ' ' || c == '\t' || c == NUL || c == 0xa0)
2889 	    return 0;	    /* blank */
2890 	if (vim_iswordc_buf(c, buf))
2891 	    return 2;	    /* word character */
2892 	return 1;	    /* punctuation */
2893     }
2894 
2895     /* binary search in table */
2896     while (top >= bot)
2897     {
2898 	mid = (bot + top) / 2;
2899 	if (classes[mid].last < (unsigned int)c)
2900 	    bot = mid + 1;
2901 	else if (classes[mid].first > (unsigned int)c)
2902 	    top = mid - 1;
2903 	else
2904 	    return (int)classes[mid].class;
2905     }
2906 
2907     /* emoji */
2908     if (intable(emoji_all, sizeof(emoji_all), c))
2909 	return 3;
2910 
2911     /* most other characters are "word" characters */
2912     return 2;
2913 }
2914 
2915     int
2916 utf_ambiguous_width(int c)
2917 {
2918     return c >= 0x80 && (intable(ambiguous, sizeof(ambiguous), c)
2919 	    || intable(emoji_all, sizeof(emoji_all), c));
2920 }
2921 
2922 /*
2923  * Code for Unicode case-dependent operations.  Based on notes in
2924  * http://www.unicode.org/Public/UNIDATA/CaseFolding.txt
2925  * This code uses simple case folding, not full case folding.
2926  * Last updated for Unicode 5.2.
2927  */
2928 
2929 /*
2930  * The following tables are built by ../runtime/tools/unicode.vim.
2931  * They must be in numeric order, because we use binary search.
2932  * An entry such as {0x41,0x5a,1,32} means that Unicode characters in the
2933  * range from 0x41 to 0x5a inclusive, stepping by 1, are changed to
2934  * folded/upper/lower by adding 32.
2935  */
2936 typedef struct
2937 {
2938     int rangeStart;
2939     int rangeEnd;
2940     int step;
2941     int offset;
2942 } convertStruct;
2943 
2944 static convertStruct foldCase[] =
2945 {
2946 	{0x41,0x5a,1,32},
2947 	{0xb5,0xb5,-1,775},
2948 	{0xc0,0xd6,1,32},
2949 	{0xd8,0xde,1,32},
2950 	{0x100,0x12e,2,1},
2951 	{0x132,0x136,2,1},
2952 	{0x139,0x147,2,1},
2953 	{0x14a,0x176,2,1},
2954 	{0x178,0x178,-1,-121},
2955 	{0x179,0x17d,2,1},
2956 	{0x17f,0x17f,-1,-268},
2957 	{0x181,0x181,-1,210},
2958 	{0x182,0x184,2,1},
2959 	{0x186,0x186,-1,206},
2960 	{0x187,0x187,-1,1},
2961 	{0x189,0x18a,1,205},
2962 	{0x18b,0x18b,-1,1},
2963 	{0x18e,0x18e,-1,79},
2964 	{0x18f,0x18f,-1,202},
2965 	{0x190,0x190,-1,203},
2966 	{0x191,0x191,-1,1},
2967 	{0x193,0x193,-1,205},
2968 	{0x194,0x194,-1,207},
2969 	{0x196,0x196,-1,211},
2970 	{0x197,0x197,-1,209},
2971 	{0x198,0x198,-1,1},
2972 	{0x19c,0x19c,-1,211},
2973 	{0x19d,0x19d,-1,213},
2974 	{0x19f,0x19f,-1,214},
2975 	{0x1a0,0x1a4,2,1},
2976 	{0x1a6,0x1a6,-1,218},
2977 	{0x1a7,0x1a7,-1,1},
2978 	{0x1a9,0x1a9,-1,218},
2979 	{0x1ac,0x1ac,-1,1},
2980 	{0x1ae,0x1ae,-1,218},
2981 	{0x1af,0x1af,-1,1},
2982 	{0x1b1,0x1b2,1,217},
2983 	{0x1b3,0x1b5,2,1},
2984 	{0x1b7,0x1b7,-1,219},
2985 	{0x1b8,0x1bc,4,1},
2986 	{0x1c4,0x1c4,-1,2},
2987 	{0x1c5,0x1c5,-1,1},
2988 	{0x1c7,0x1c7,-1,2},
2989 	{0x1c8,0x1c8,-1,1},
2990 	{0x1ca,0x1ca,-1,2},
2991 	{0x1cb,0x1db,2,1},
2992 	{0x1de,0x1ee,2,1},
2993 	{0x1f1,0x1f1,-1,2},
2994 	{0x1f2,0x1f4,2,1},
2995 	{0x1f6,0x1f6,-1,-97},
2996 	{0x1f7,0x1f7,-1,-56},
2997 	{0x1f8,0x21e,2,1},
2998 	{0x220,0x220,-1,-130},
2999 	{0x222,0x232,2,1},
3000 	{0x23a,0x23a,-1,10795},
3001 	{0x23b,0x23b,-1,1},
3002 	{0x23d,0x23d,-1,-163},
3003 	{0x23e,0x23e,-1,10792},
3004 	{0x241,0x241,-1,1},
3005 	{0x243,0x243,-1,-195},
3006 	{0x244,0x244,-1,69},
3007 	{0x245,0x245,-1,71},
3008 	{0x246,0x24e,2,1},
3009 	{0x345,0x345,-1,116},
3010 	{0x370,0x372,2,1},
3011 	{0x376,0x376,-1,1},
3012 	{0x37f,0x37f,-1,116},
3013 	{0x386,0x386,-1,38},
3014 	{0x388,0x38a,1,37},
3015 	{0x38c,0x38c,-1,64},
3016 	{0x38e,0x38f,1,63},
3017 	{0x391,0x3a1,1,32},
3018 	{0x3a3,0x3ab,1,32},
3019 	{0x3c2,0x3c2,-1,1},
3020 	{0x3cf,0x3cf,-1,8},
3021 	{0x3d0,0x3d0,-1,-30},
3022 	{0x3d1,0x3d1,-1,-25},
3023 	{0x3d5,0x3d5,-1,-15},
3024 	{0x3d6,0x3d6,-1,-22},
3025 	{0x3d8,0x3ee,2,1},
3026 	{0x3f0,0x3f0,-1,-54},
3027 	{0x3f1,0x3f1,-1,-48},
3028 	{0x3f4,0x3f4,-1,-60},
3029 	{0x3f5,0x3f5,-1,-64},
3030 	{0x3f7,0x3f7,-1,1},
3031 	{0x3f9,0x3f9,-1,-7},
3032 	{0x3fa,0x3fa,-1,1},
3033 	{0x3fd,0x3ff,1,-130},
3034 	{0x400,0x40f,1,80},
3035 	{0x410,0x42f,1,32},
3036 	{0x460,0x480,2,1},
3037 	{0x48a,0x4be,2,1},
3038 	{0x4c0,0x4c0,-1,15},
3039 	{0x4c1,0x4cd,2,1},
3040 	{0x4d0,0x52e,2,1},
3041 	{0x531,0x556,1,48},
3042 	{0x10a0,0x10c5,1,7264},
3043 	{0x10c7,0x10cd,6,7264},
3044 	{0x13f8,0x13fd,1,-8},
3045 	{0x1c80,0x1c80,-1,-6222},
3046 	{0x1c81,0x1c81,-1,-6221},
3047 	{0x1c82,0x1c82,-1,-6212},
3048 	{0x1c83,0x1c84,1,-6210},
3049 	{0x1c85,0x1c85,-1,-6211},
3050 	{0x1c86,0x1c86,-1,-6204},
3051 	{0x1c87,0x1c87,-1,-6180},
3052 	{0x1c88,0x1c88,-1,35267},
3053 	{0x1c90,0x1cba,1,-3008},
3054 	{0x1cbd,0x1cbf,1,-3008},
3055 	{0x1e00,0x1e94,2,1},
3056 	{0x1e9b,0x1e9b,-1,-58},
3057 	{0x1e9e,0x1e9e,-1,-7615},
3058 	{0x1ea0,0x1efe,2,1},
3059 	{0x1f08,0x1f0f,1,-8},
3060 	{0x1f18,0x1f1d,1,-8},
3061 	{0x1f28,0x1f2f,1,-8},
3062 	{0x1f38,0x1f3f,1,-8},
3063 	{0x1f48,0x1f4d,1,-8},
3064 	{0x1f59,0x1f5f,2,-8},
3065 	{0x1f68,0x1f6f,1,-8},
3066 	{0x1f88,0x1f8f,1,-8},
3067 	{0x1f98,0x1f9f,1,-8},
3068 	{0x1fa8,0x1faf,1,-8},
3069 	{0x1fb8,0x1fb9,1,-8},
3070 	{0x1fba,0x1fbb,1,-74},
3071 	{0x1fbc,0x1fbc,-1,-9},
3072 	{0x1fbe,0x1fbe,-1,-7173},
3073 	{0x1fc8,0x1fcb,1,-86},
3074 	{0x1fcc,0x1fcc,-1,-9},
3075 	{0x1fd8,0x1fd9,1,-8},
3076 	{0x1fda,0x1fdb,1,-100},
3077 	{0x1fe8,0x1fe9,1,-8},
3078 	{0x1fea,0x1feb,1,-112},
3079 	{0x1fec,0x1fec,-1,-7},
3080 	{0x1ff8,0x1ff9,1,-128},
3081 	{0x1ffa,0x1ffb,1,-126},
3082 	{0x1ffc,0x1ffc,-1,-9},
3083 	{0x2126,0x2126,-1,-7517},
3084 	{0x212a,0x212a,-1,-8383},
3085 	{0x212b,0x212b,-1,-8262},
3086 	{0x2132,0x2132,-1,28},
3087 	{0x2160,0x216f,1,16},
3088 	{0x2183,0x2183,-1,1},
3089 	{0x24b6,0x24cf,1,26},
3090 	{0x2c00,0x2c2e,1,48},
3091 	{0x2c60,0x2c60,-1,1},
3092 	{0x2c62,0x2c62,-1,-10743},
3093 	{0x2c63,0x2c63,-1,-3814},
3094 	{0x2c64,0x2c64,-1,-10727},
3095 	{0x2c67,0x2c6b,2,1},
3096 	{0x2c6d,0x2c6d,-1,-10780},
3097 	{0x2c6e,0x2c6e,-1,-10749},
3098 	{0x2c6f,0x2c6f,-1,-10783},
3099 	{0x2c70,0x2c70,-1,-10782},
3100 	{0x2c72,0x2c75,3,1},
3101 	{0x2c7e,0x2c7f,1,-10815},
3102 	{0x2c80,0x2ce2,2,1},
3103 	{0x2ceb,0x2ced,2,1},
3104 	{0x2cf2,0xa640,31054,1},
3105 	{0xa642,0xa66c,2,1},
3106 	{0xa680,0xa69a,2,1},
3107 	{0xa722,0xa72e,2,1},
3108 	{0xa732,0xa76e,2,1},
3109 	{0xa779,0xa77b,2,1},
3110 	{0xa77d,0xa77d,-1,-35332},
3111 	{0xa77e,0xa786,2,1},
3112 	{0xa78b,0xa78b,-1,1},
3113 	{0xa78d,0xa78d,-1,-42280},
3114 	{0xa790,0xa792,2,1},
3115 	{0xa796,0xa7a8,2,1},
3116 	{0xa7aa,0xa7aa,-1,-42308},
3117 	{0xa7ab,0xa7ab,-1,-42319},
3118 	{0xa7ac,0xa7ac,-1,-42315},
3119 	{0xa7ad,0xa7ad,-1,-42305},
3120 	{0xa7ae,0xa7ae,-1,-42308},
3121 	{0xa7b0,0xa7b0,-1,-42258},
3122 	{0xa7b1,0xa7b1,-1,-42282},
3123 	{0xa7b2,0xa7b2,-1,-42261},
3124 	{0xa7b3,0xa7b3,-1,928},
3125 	{0xa7b4,0xa7b8,2,1},
3126 	{0xab70,0xabbf,1,-38864},
3127 	{0xff21,0xff3a,1,32},
3128 	{0x10400,0x10427,1,40},
3129 	{0x104b0,0x104d3,1,40},
3130 	{0x10c80,0x10cb2,1,64},
3131 	{0x118a0,0x118bf,1,32},
3132 	{0x16e40,0x16e5f,1,32},
3133 	{0x1e900,0x1e921,1,34}
3134 };
3135 
3136 static int utf_convert(int a, convertStruct table[], int tableSize);
3137 static int utf_strnicmp(char_u *s1, char_u *s2, size_t n1, size_t n2);
3138 
3139 /*
3140  * Generic conversion function for case operations.
3141  * Return the converted equivalent of "a", which is a UCS-4 character.  Use
3142  * the given conversion "table".  Uses binary search on "table".
3143  */
3144     static int
3145 utf_convert(
3146     int			a,
3147     convertStruct	table[],
3148     int			tableSize)
3149 {
3150     int start, mid, end; /* indices into table */
3151     int entries = tableSize / sizeof(convertStruct);
3152 
3153     start = 0;
3154     end = entries;
3155     while (start < end)
3156     {
3157 	/* need to search further */
3158 	mid = (end + start) / 2;
3159 	if (table[mid].rangeEnd < a)
3160 	    start = mid + 1;
3161 	else
3162 	    end = mid;
3163     }
3164     if (start < entries
3165 	    && table[start].rangeStart <= a
3166 	    && a <= table[start].rangeEnd
3167 	    && (a - table[start].rangeStart) % table[start].step == 0)
3168 	return (a + table[start].offset);
3169     else
3170 	return a;
3171 }
3172 
3173 /*
3174  * Return the folded-case equivalent of "a", which is a UCS-4 character.  Uses
3175  * simple case folding.
3176  */
3177     int
3178 utf_fold(int a)
3179 {
3180     if (a < 0x80)
3181 	/* be fast for ASCII */
3182 	return a >= 0x41 && a <= 0x5a ? a + 32 : a;
3183     return utf_convert(a, foldCase, (int)sizeof(foldCase));
3184 }
3185 
3186 static convertStruct toLower[] =
3187 {
3188 	{0x41,0x5a,1,32},
3189 	{0xc0,0xd6,1,32},
3190 	{0xd8,0xde,1,32},
3191 	{0x100,0x12e,2,1},
3192 	{0x130,0x130,-1,-199},
3193 	{0x132,0x136,2,1},
3194 	{0x139,0x147,2,1},
3195 	{0x14a,0x176,2,1},
3196 	{0x178,0x178,-1,-121},
3197 	{0x179,0x17d,2,1},
3198 	{0x181,0x181,-1,210},
3199 	{0x182,0x184,2,1},
3200 	{0x186,0x186,-1,206},
3201 	{0x187,0x187,-1,1},
3202 	{0x189,0x18a,1,205},
3203 	{0x18b,0x18b,-1,1},
3204 	{0x18e,0x18e,-1,79},
3205 	{0x18f,0x18f,-1,202},
3206 	{0x190,0x190,-1,203},
3207 	{0x191,0x191,-1,1},
3208 	{0x193,0x193,-1,205},
3209 	{0x194,0x194,-1,207},
3210 	{0x196,0x196,-1,211},
3211 	{0x197,0x197,-1,209},
3212 	{0x198,0x198,-1,1},
3213 	{0x19c,0x19c,-1,211},
3214 	{0x19d,0x19d,-1,213},
3215 	{0x19f,0x19f,-1,214},
3216 	{0x1a0,0x1a4,2,1},
3217 	{0x1a6,0x1a6,-1,218},
3218 	{0x1a7,0x1a7,-1,1},
3219 	{0x1a9,0x1a9,-1,218},
3220 	{0x1ac,0x1ac,-1,1},
3221 	{0x1ae,0x1ae,-1,218},
3222 	{0x1af,0x1af,-1,1},
3223 	{0x1b1,0x1b2,1,217},
3224 	{0x1b3,0x1b5,2,1},
3225 	{0x1b7,0x1b7,-1,219},
3226 	{0x1b8,0x1bc,4,1},
3227 	{0x1c4,0x1c4,-1,2},
3228 	{0x1c5,0x1c5,-1,1},
3229 	{0x1c7,0x1c7,-1,2},
3230 	{0x1c8,0x1c8,-1,1},
3231 	{0x1ca,0x1ca,-1,2},
3232 	{0x1cb,0x1db,2,1},
3233 	{0x1de,0x1ee,2,1},
3234 	{0x1f1,0x1f1,-1,2},
3235 	{0x1f2,0x1f4,2,1},
3236 	{0x1f6,0x1f6,-1,-97},
3237 	{0x1f7,0x1f7,-1,-56},
3238 	{0x1f8,0x21e,2,1},
3239 	{0x220,0x220,-1,-130},
3240 	{0x222,0x232,2,1},
3241 	{0x23a,0x23a,-1,10795},
3242 	{0x23b,0x23b,-1,1},
3243 	{0x23d,0x23d,-1,-163},
3244 	{0x23e,0x23e,-1,10792},
3245 	{0x241,0x241,-1,1},
3246 	{0x243,0x243,-1,-195},
3247 	{0x244,0x244,-1,69},
3248 	{0x245,0x245,-1,71},
3249 	{0x246,0x24e,2,1},
3250 	{0x370,0x372,2,1},
3251 	{0x376,0x376,-1,1},
3252 	{0x37f,0x37f,-1,116},
3253 	{0x386,0x386,-1,38},
3254 	{0x388,0x38a,1,37},
3255 	{0x38c,0x38c,-1,64},
3256 	{0x38e,0x38f,1,63},
3257 	{0x391,0x3a1,1,32},
3258 	{0x3a3,0x3ab,1,32},
3259 	{0x3cf,0x3cf,-1,8},
3260 	{0x3d8,0x3ee,2,1},
3261 	{0x3f4,0x3f4,-1,-60},
3262 	{0x3f7,0x3f7,-1,1},
3263 	{0x3f9,0x3f9,-1,-7},
3264 	{0x3fa,0x3fa,-1,1},
3265 	{0x3fd,0x3ff,1,-130},
3266 	{0x400,0x40f,1,80},
3267 	{0x410,0x42f,1,32},
3268 	{0x460,0x480,2,1},
3269 	{0x48a,0x4be,2,1},
3270 	{0x4c0,0x4c0,-1,15},
3271 	{0x4c1,0x4cd,2,1},
3272 	{0x4d0,0x52e,2,1},
3273 	{0x531,0x556,1,48},
3274 	{0x10a0,0x10c5,1,7264},
3275 	{0x10c7,0x10cd,6,7264},
3276 	{0x13a0,0x13ef,1,38864},
3277 	{0x13f0,0x13f5,1,8},
3278 	{0x1c90,0x1cba,1,-3008},
3279 	{0x1cbd,0x1cbf,1,-3008},
3280 	{0x1e00,0x1e94,2,1},
3281 	{0x1e9e,0x1e9e,-1,-7615},
3282 	{0x1ea0,0x1efe,2,1},
3283 	{0x1f08,0x1f0f,1,-8},
3284 	{0x1f18,0x1f1d,1,-8},
3285 	{0x1f28,0x1f2f,1,-8},
3286 	{0x1f38,0x1f3f,1,-8},
3287 	{0x1f48,0x1f4d,1,-8},
3288 	{0x1f59,0x1f5f,2,-8},
3289 	{0x1f68,0x1f6f,1,-8},
3290 	{0x1f88,0x1f8f,1,-8},
3291 	{0x1f98,0x1f9f,1,-8},
3292 	{0x1fa8,0x1faf,1,-8},
3293 	{0x1fb8,0x1fb9,1,-8},
3294 	{0x1fba,0x1fbb,1,-74},
3295 	{0x1fbc,0x1fbc,-1,-9},
3296 	{0x1fc8,0x1fcb,1,-86},
3297 	{0x1fcc,0x1fcc,-1,-9},
3298 	{0x1fd8,0x1fd9,1,-8},
3299 	{0x1fda,0x1fdb,1,-100},
3300 	{0x1fe8,0x1fe9,1,-8},
3301 	{0x1fea,0x1feb,1,-112},
3302 	{0x1fec,0x1fec,-1,-7},
3303 	{0x1ff8,0x1ff9,1,-128},
3304 	{0x1ffa,0x1ffb,1,-126},
3305 	{0x1ffc,0x1ffc,-1,-9},
3306 	{0x2126,0x2126,-1,-7517},
3307 	{0x212a,0x212a,-1,-8383},
3308 	{0x212b,0x212b,-1,-8262},
3309 	{0x2132,0x2132,-1,28},
3310 	{0x2160,0x216f,1,16},
3311 	{0x2183,0x2183,-1,1},
3312 	{0x24b6,0x24cf,1,26},
3313 	{0x2c00,0x2c2e,1,48},
3314 	{0x2c60,0x2c60,-1,1},
3315 	{0x2c62,0x2c62,-1,-10743},
3316 	{0x2c63,0x2c63,-1,-3814},
3317 	{0x2c64,0x2c64,-1,-10727},
3318 	{0x2c67,0x2c6b,2,1},
3319 	{0x2c6d,0x2c6d,-1,-10780},
3320 	{0x2c6e,0x2c6e,-1,-10749},
3321 	{0x2c6f,0x2c6f,-1,-10783},
3322 	{0x2c70,0x2c70,-1,-10782},
3323 	{0x2c72,0x2c75,3,1},
3324 	{0x2c7e,0x2c7f,1,-10815},
3325 	{0x2c80,0x2ce2,2,1},
3326 	{0x2ceb,0x2ced,2,1},
3327 	{0x2cf2,0xa640,31054,1},
3328 	{0xa642,0xa66c,2,1},
3329 	{0xa680,0xa69a,2,1},
3330 	{0xa722,0xa72e,2,1},
3331 	{0xa732,0xa76e,2,1},
3332 	{0xa779,0xa77b,2,1},
3333 	{0xa77d,0xa77d,-1,-35332},
3334 	{0xa77e,0xa786,2,1},
3335 	{0xa78b,0xa78b,-1,1},
3336 	{0xa78d,0xa78d,-1,-42280},
3337 	{0xa790,0xa792,2,1},
3338 	{0xa796,0xa7a8,2,1},
3339 	{0xa7aa,0xa7aa,-1,-42308},
3340 	{0xa7ab,0xa7ab,-1,-42319},
3341 	{0xa7ac,0xa7ac,-1,-42315},
3342 	{0xa7ad,0xa7ad,-1,-42305},
3343 	{0xa7ae,0xa7ae,-1,-42308},
3344 	{0xa7b0,0xa7b0,-1,-42258},
3345 	{0xa7b1,0xa7b1,-1,-42282},
3346 	{0xa7b2,0xa7b2,-1,-42261},
3347 	{0xa7b3,0xa7b3,-1,928},
3348 	{0xa7b4,0xa7b8,2,1},
3349 	{0xff21,0xff3a,1,32},
3350 	{0x10400,0x10427,1,40},
3351 	{0x104b0,0x104d3,1,40},
3352 	{0x10c80,0x10cb2,1,64},
3353 	{0x118a0,0x118bf,1,32},
3354 	{0x16e40,0x16e5f,1,32},
3355 	{0x1e900,0x1e921,1,34}
3356 };
3357 
3358 static convertStruct toUpper[] =
3359 {
3360 	{0x61,0x7a,1,-32},
3361 	{0xb5,0xb5,-1,743},
3362 	{0xe0,0xf6,1,-32},
3363 	{0xf8,0xfe,1,-32},
3364 	{0xff,0xff,-1,121},
3365 	{0x101,0x12f,2,-1},
3366 	{0x131,0x131,-1,-232},
3367 	{0x133,0x137,2,-1},
3368 	{0x13a,0x148,2,-1},
3369 	{0x14b,0x177,2,-1},
3370 	{0x17a,0x17e,2,-1},
3371 	{0x17f,0x17f,-1,-300},
3372 	{0x180,0x180,-1,195},
3373 	{0x183,0x185,2,-1},
3374 	{0x188,0x18c,4,-1},
3375 	{0x192,0x192,-1,-1},
3376 	{0x195,0x195,-1,97},
3377 	{0x199,0x199,-1,-1},
3378 	{0x19a,0x19a,-1,163},
3379 	{0x19e,0x19e,-1,130},
3380 	{0x1a1,0x1a5,2,-1},
3381 	{0x1a8,0x1ad,5,-1},
3382 	{0x1b0,0x1b4,4,-1},
3383 	{0x1b6,0x1b9,3,-1},
3384 	{0x1bd,0x1bd,-1,-1},
3385 	{0x1bf,0x1bf,-1,56},
3386 	{0x1c5,0x1c5,-1,-1},
3387 	{0x1c6,0x1c6,-1,-2},
3388 	{0x1c8,0x1c8,-1,-1},
3389 	{0x1c9,0x1c9,-1,-2},
3390 	{0x1cb,0x1cb,-1,-1},
3391 	{0x1cc,0x1cc,-1,-2},
3392 	{0x1ce,0x1dc,2,-1},
3393 	{0x1dd,0x1dd,-1,-79},
3394 	{0x1df,0x1ef,2,-1},
3395 	{0x1f2,0x1f2,-1,-1},
3396 	{0x1f3,0x1f3,-1,-2},
3397 	{0x1f5,0x1f9,4,-1},
3398 	{0x1fb,0x21f,2,-1},
3399 	{0x223,0x233,2,-1},
3400 	{0x23c,0x23c,-1,-1},
3401 	{0x23f,0x240,1,10815},
3402 	{0x242,0x247,5,-1},
3403 	{0x249,0x24f,2,-1},
3404 	{0x250,0x250,-1,10783},
3405 	{0x251,0x251,-1,10780},
3406 	{0x252,0x252,-1,10782},
3407 	{0x253,0x253,-1,-210},
3408 	{0x254,0x254,-1,-206},
3409 	{0x256,0x257,1,-205},
3410 	{0x259,0x259,-1,-202},
3411 	{0x25b,0x25b,-1,-203},
3412 	{0x25c,0x25c,-1,42319},
3413 	{0x260,0x260,-1,-205},
3414 	{0x261,0x261,-1,42315},
3415 	{0x263,0x263,-1,-207},
3416 	{0x265,0x265,-1,42280},
3417 	{0x266,0x266,-1,42308},
3418 	{0x268,0x268,-1,-209},
3419 	{0x269,0x269,-1,-211},
3420 	{0x26a,0x26a,-1,42308},
3421 	{0x26b,0x26b,-1,10743},
3422 	{0x26c,0x26c,-1,42305},
3423 	{0x26f,0x26f,-1,-211},
3424 	{0x271,0x271,-1,10749},
3425 	{0x272,0x272,-1,-213},
3426 	{0x275,0x275,-1,-214},
3427 	{0x27d,0x27d,-1,10727},
3428 	{0x280,0x283,3,-218},
3429 	{0x287,0x287,-1,42282},
3430 	{0x288,0x288,-1,-218},
3431 	{0x289,0x289,-1,-69},
3432 	{0x28a,0x28b,1,-217},
3433 	{0x28c,0x28c,-1,-71},
3434 	{0x292,0x292,-1,-219},
3435 	{0x29d,0x29d,-1,42261},
3436 	{0x29e,0x29e,-1,42258},
3437 	{0x345,0x345,-1,84},
3438 	{0x371,0x373,2,-1},
3439 	{0x377,0x377,-1,-1},
3440 	{0x37b,0x37d,1,130},
3441 	{0x3ac,0x3ac,-1,-38},
3442 	{0x3ad,0x3af,1,-37},
3443 	{0x3b1,0x3c1,1,-32},
3444 	{0x3c2,0x3c2,-1,-31},
3445 	{0x3c3,0x3cb,1,-32},
3446 	{0x3cc,0x3cc,-1,-64},
3447 	{0x3cd,0x3ce,1,-63},
3448 	{0x3d0,0x3d0,-1,-62},
3449 	{0x3d1,0x3d1,-1,-57},
3450 	{0x3d5,0x3d5,-1,-47},
3451 	{0x3d6,0x3d6,-1,-54},
3452 	{0x3d7,0x3d7,-1,-8},
3453 	{0x3d9,0x3ef,2,-1},
3454 	{0x3f0,0x3f0,-1,-86},
3455 	{0x3f1,0x3f1,-1,-80},
3456 	{0x3f2,0x3f2,-1,7},
3457 	{0x3f3,0x3f3,-1,-116},
3458 	{0x3f5,0x3f5,-1,-96},
3459 	{0x3f8,0x3fb,3,-1},
3460 	{0x430,0x44f,1,-32},
3461 	{0x450,0x45f,1,-80},
3462 	{0x461,0x481,2,-1},
3463 	{0x48b,0x4bf,2,-1},
3464 	{0x4c2,0x4ce,2,-1},
3465 	{0x4cf,0x4cf,-1,-15},
3466 	{0x4d1,0x52f,2,-1},
3467 	{0x561,0x586,1,-48},
3468 	{0x10d0,0x10fa,1,3008},
3469 	{0x10fd,0x10ff,1,3008},
3470 	{0x13f8,0x13fd,1,-8},
3471 	{0x1c80,0x1c80,-1,-6254},
3472 	{0x1c81,0x1c81,-1,-6253},
3473 	{0x1c82,0x1c82,-1,-6244},
3474 	{0x1c83,0x1c84,1,-6242},
3475 	{0x1c85,0x1c85,-1,-6243},
3476 	{0x1c86,0x1c86,-1,-6236},
3477 	{0x1c87,0x1c87,-1,-6181},
3478 	{0x1c88,0x1c88,-1,35266},
3479 	{0x1d79,0x1d79,-1,35332},
3480 	{0x1d7d,0x1d7d,-1,3814},
3481 	{0x1e01,0x1e95,2,-1},
3482 	{0x1e9b,0x1e9b,-1,-59},
3483 	{0x1ea1,0x1eff,2,-1},
3484 	{0x1f00,0x1f07,1,8},
3485 	{0x1f10,0x1f15,1,8},
3486 	{0x1f20,0x1f27,1,8},
3487 	{0x1f30,0x1f37,1,8},
3488 	{0x1f40,0x1f45,1,8},
3489 	{0x1f51,0x1f57,2,8},
3490 	{0x1f60,0x1f67,1,8},
3491 	{0x1f70,0x1f71,1,74},
3492 	{0x1f72,0x1f75,1,86},
3493 	{0x1f76,0x1f77,1,100},
3494 	{0x1f78,0x1f79,1,128},
3495 	{0x1f7a,0x1f7b,1,112},
3496 	{0x1f7c,0x1f7d,1,126},
3497 	{0x1f80,0x1f87,1,8},
3498 	{0x1f90,0x1f97,1,8},
3499 	{0x1fa0,0x1fa7,1,8},
3500 	{0x1fb0,0x1fb1,1,8},
3501 	{0x1fb3,0x1fb3,-1,9},
3502 	{0x1fbe,0x1fbe,-1,-7205},
3503 	{0x1fc3,0x1fc3,-1,9},
3504 	{0x1fd0,0x1fd1,1,8},
3505 	{0x1fe0,0x1fe1,1,8},
3506 	{0x1fe5,0x1fe5,-1,7},
3507 	{0x1ff3,0x1ff3,-1,9},
3508 	{0x214e,0x214e,-1,-28},
3509 	{0x2170,0x217f,1,-16},
3510 	{0x2184,0x2184,-1,-1},
3511 	{0x24d0,0x24e9,1,-26},
3512 	{0x2c30,0x2c5e,1,-48},
3513 	{0x2c61,0x2c61,-1,-1},
3514 	{0x2c65,0x2c65,-1,-10795},
3515 	{0x2c66,0x2c66,-1,-10792},
3516 	{0x2c68,0x2c6c,2,-1},
3517 	{0x2c73,0x2c76,3,-1},
3518 	{0x2c81,0x2ce3,2,-1},
3519 	{0x2cec,0x2cee,2,-1},
3520 	{0x2cf3,0x2cf3,-1,-1},
3521 	{0x2d00,0x2d25,1,-7264},
3522 	{0x2d27,0x2d2d,6,-7264},
3523 	{0xa641,0xa66d,2,-1},
3524 	{0xa681,0xa69b,2,-1},
3525 	{0xa723,0xa72f,2,-1},
3526 	{0xa733,0xa76f,2,-1},
3527 	{0xa77a,0xa77c,2,-1},
3528 	{0xa77f,0xa787,2,-1},
3529 	{0xa78c,0xa791,5,-1},
3530 	{0xa793,0xa797,4,-1},
3531 	{0xa799,0xa7a9,2,-1},
3532 	{0xa7b5,0xa7b9,2,-1},
3533 	{0xab53,0xab53,-1,-928},
3534 	{0xab70,0xabbf,1,-38864},
3535 	{0xff41,0xff5a,1,-32},
3536 	{0x10428,0x1044f,1,-40},
3537 	{0x104d8,0x104fb,1,-40},
3538 	{0x10cc0,0x10cf2,1,-64},
3539 	{0x118c0,0x118df,1,-32},
3540 	{0x16e60,0x16e7f,1,-32},
3541 	{0x1e922,0x1e943,1,-34}
3542 };
3543 
3544 /*
3545  * Return the upper-case equivalent of "a", which is a UCS-4 character.  Use
3546  * simple case folding.
3547  */
3548     int
3549 utf_toupper(int a)
3550 {
3551     /* If 'casemap' contains "keepascii" use ASCII style toupper(). */
3552     if (a < 128 && (cmp_flags & CMP_KEEPASCII))
3553 	return TOUPPER_ASC(a);
3554 
3555 #if defined(HAVE_TOWUPPER) && defined(__STDC_ISO_10646__)
3556     /* If towupper() is available and handles Unicode, use it. */
3557     if (!(cmp_flags & CMP_INTERNAL))
3558 	return towupper(a);
3559 #endif
3560 
3561     /* For characters below 128 use locale sensitive toupper(). */
3562     if (a < 128)
3563 	return TOUPPER_LOC(a);
3564 
3565     /* For any other characters use the above mapping table. */
3566     return utf_convert(a, toUpper, (int)sizeof(toUpper));
3567 }
3568 
3569     int
3570 utf_islower(int a)
3571 {
3572     /* German sharp s is lower case but has no upper case equivalent. */
3573     return (utf_toupper(a) != a) || a == 0xdf;
3574 }
3575 
3576 /*
3577  * Return the lower-case equivalent of "a", which is a UCS-4 character.  Use
3578  * simple case folding.
3579  */
3580     int
3581 utf_tolower(int a)
3582 {
3583     /* If 'casemap' contains "keepascii" use ASCII style tolower(). */
3584     if (a < 128 && (cmp_flags & CMP_KEEPASCII))
3585 	return TOLOWER_ASC(a);
3586 
3587 #if defined(HAVE_TOWLOWER) && defined(__STDC_ISO_10646__)
3588     /* If towlower() is available and handles Unicode, use it. */
3589     if (!(cmp_flags & CMP_INTERNAL))
3590 	return towlower(a);
3591 #endif
3592 
3593     /* For characters below 128 use locale sensitive tolower(). */
3594     if (a < 128)
3595 	return TOLOWER_LOC(a);
3596 
3597     /* For any other characters use the above mapping table. */
3598     return utf_convert(a, toLower, (int)sizeof(toLower));
3599 }
3600 
3601     int
3602 utf_isupper(int a)
3603 {
3604     return (utf_tolower(a) != a);
3605 }
3606 
3607     static int
3608 utf_strnicmp(
3609     char_u      *s1,
3610     char_u      *s2,
3611     size_t      n1,
3612     size_t      n2)
3613 {
3614     int		c1, c2, cdiff;
3615     char_u	buffer[6];
3616 
3617     for (;;)
3618     {
3619 	c1 = utf_safe_read_char_adv(&s1, &n1);
3620 	c2 = utf_safe_read_char_adv(&s2, &n2);
3621 
3622 	if (c1 <= 0 || c2 <= 0)
3623 	    break;
3624 
3625 	if (c1 == c2)
3626 	    continue;
3627 
3628 	cdiff = utf_fold(c1) - utf_fold(c2);
3629 	if (cdiff != 0)
3630 	    return cdiff;
3631     }
3632 
3633     /* some string ended or has an incomplete/illegal character sequence */
3634 
3635     if (c1 == 0 || c2 == 0)
3636     {
3637 	/* some string ended. shorter string is smaller */
3638 	if (c1 == 0 && c2 == 0)
3639 	    return 0;
3640 	return c1 == 0 ? -1 : 1;
3641     }
3642 
3643     /* Continue with bytewise comparison to produce some result that
3644      * would make comparison operations involving this function transitive.
3645      *
3646      * If only one string had an error, comparison should be made with
3647      * folded version of the other string. In this case it is enough
3648      * to fold just one character to determine the result of comparison. */
3649 
3650     if (c1 != -1 && c2 == -1)
3651     {
3652 	n1 = utf_char2bytes(utf_fold(c1), buffer);
3653 	s1 = buffer;
3654     }
3655     else if (c2 != -1 && c1 == -1)
3656     {
3657 	n2 = utf_char2bytes(utf_fold(c2), buffer);
3658 	s2 = buffer;
3659     }
3660 
3661     while (n1 > 0 && n2 > 0 && *s1 != NUL && *s2 != NUL)
3662     {
3663 	cdiff = (int)(*s1) - (int)(*s2);
3664 	if (cdiff != 0)
3665 	    return cdiff;
3666 
3667 	s1++;
3668 	s2++;
3669 	n1--;
3670 	n2--;
3671     }
3672 
3673     if (n1 > 0 && *s1 == NUL)
3674 	n1 = 0;
3675     if (n2 > 0 && *s2 == NUL)
3676 	n2 = 0;
3677 
3678     if (n1 == 0 && n2 == 0)
3679 	return 0;
3680     return n1 == 0 ? -1 : 1;
3681 }
3682 
3683 /*
3684  * Version of strnicmp() that handles multi-byte characters.
3685  * Needed for Big5, Shift-JIS and UTF-8 encoding.  Other DBCS encodings can
3686  * probably use strnicmp(), because there are no ASCII characters in the
3687  * second byte.
3688  * Returns zero if s1 and s2 are equal (ignoring case), the difference between
3689  * two characters otherwise.
3690  */
3691     int
3692 mb_strnicmp(char_u *s1, char_u *s2, size_t nn)
3693 {
3694     int		i, l;
3695     int		cdiff;
3696     int		n = (int)nn;
3697 
3698     if (enc_utf8)
3699     {
3700 	return utf_strnicmp(s1, s2, nn, nn);
3701     }
3702     else
3703     {
3704 	for (i = 0; i < n; i += l)
3705 	{
3706 	    if (s1[i] == NUL && s2[i] == NUL)	/* both strings end */
3707 		return 0;
3708 
3709 	    l = (*mb_ptr2len)(s1 + i);
3710 	    if (l <= 1)
3711 	    {
3712 		/* Single byte: first check normally, then with ignore case. */
3713 		if (s1[i] != s2[i])
3714 		{
3715 		    cdiff = MB_TOLOWER(s1[i]) - MB_TOLOWER(s2[i]);
3716 		    if (cdiff != 0)
3717 			return cdiff;
3718 		}
3719 	    }
3720 	    else
3721 	    {
3722 		/* For non-Unicode multi-byte don't ignore case. */
3723 		if (l > n - i)
3724 		    l = n - i;
3725 		cdiff = STRNCMP(s1 + i, s2 + i, l);
3726 		if (cdiff != 0)
3727 		    return cdiff;
3728 	    }
3729 	}
3730     }
3731     return 0;
3732 }
3733 
3734 /*
3735  * "g8": show bytes of the UTF-8 char under the cursor.  Doesn't matter what
3736  * 'encoding' has been set to.
3737  */
3738     void
3739 show_utf8(void)
3740 {
3741     int		len;
3742     int		rlen = 0;
3743     char_u	*line;
3744     int		clen;
3745     int		i;
3746 
3747     /* Get the byte length of the char under the cursor, including composing
3748      * characters. */
3749     line = ml_get_cursor();
3750     len = utfc_ptr2len(line);
3751     if (len == 0)
3752     {
3753 	MSG("NUL");
3754 	return;
3755     }
3756 
3757     clen = 0;
3758     for (i = 0; i < len; ++i)
3759     {
3760 	if (clen == 0)
3761 	{
3762 	    /* start of (composing) character, get its length */
3763 	    if (i > 0)
3764 	    {
3765 		STRCPY(IObuff + rlen, "+ ");
3766 		rlen += 2;
3767 	    }
3768 	    clen = utf_ptr2len(line + i);
3769 	}
3770 	sprintf((char *)IObuff + rlen, "%02x ",
3771 		(line[i] == NL) ? NUL : line[i]);  /* NUL is stored as NL */
3772 	--clen;
3773 	rlen += (int)STRLEN(IObuff + rlen);
3774 	if (rlen > IOSIZE - 20)
3775 	    break;
3776     }
3777 
3778     msg(IObuff);
3779 }
3780 
3781 /*
3782  * mb_head_off() function pointer.
3783  * Return offset from "p" to the first byte of the character it points into.
3784  * If "p" points to the NUL at the end of the string return 0.
3785  * Returns 0 when already at the first byte of a character.
3786  */
3787     int
3788 latin_head_off(char_u *base UNUSED, char_u *p UNUSED)
3789 {
3790     return 0;
3791 }
3792 
3793     int
3794 dbcs_head_off(char_u *base, char_u *p)
3795 {
3796     char_u	*q;
3797 
3798     /* It can't be a trailing byte when not using DBCS, at the start of the
3799      * string or the previous byte can't start a double-byte. */
3800     if (p <= base || MB_BYTE2LEN(p[-1]) == 1 || *p == NUL)
3801 	return 0;
3802 
3803     /* This is slow: need to start at the base and go forward until the
3804      * byte we are looking for.  Return 1 when we went past it, 0 otherwise. */
3805     q = base;
3806     while (q < p)
3807 	q += dbcs_ptr2len(q);
3808     return (q == p) ? 0 : 1;
3809 }
3810 
3811 /*
3812  * Special version of dbcs_head_off() that works for ScreenLines[], where
3813  * single-width DBCS_JPNU characters are stored separately.
3814  */
3815     int
3816 dbcs_screen_head_off(char_u *base, char_u *p)
3817 {
3818     char_u	*q;
3819 
3820     /* It can't be a trailing byte when not using DBCS, at the start of the
3821      * string or the previous byte can't start a double-byte.
3822      * For euc-jp an 0x8e byte in the previous cell always means we have a
3823      * lead byte in the current cell. */
3824     if (p <= base
3825 	    || (enc_dbcs == DBCS_JPNU && p[-1] == 0x8e)
3826 	    || MB_BYTE2LEN(p[-1]) == 1
3827 	    || *p == NUL)
3828 	return 0;
3829 
3830     /* This is slow: need to start at the base and go forward until the
3831      * byte we are looking for.  Return 1 when we went past it, 0 otherwise.
3832      * For DBCS_JPNU look out for 0x8e, which means the second byte is not
3833      * stored as the next byte. */
3834     q = base;
3835     while (q < p)
3836     {
3837 	if (enc_dbcs == DBCS_JPNU && *q == 0x8e)
3838 	    ++q;
3839 	else
3840 	    q += dbcs_ptr2len(q);
3841     }
3842     return (q == p) ? 0 : 1;
3843 }
3844 
3845     int
3846 utf_head_off(char_u *base, char_u *p)
3847 {
3848     char_u	*q;
3849     char_u	*s;
3850     int		c;
3851     int		len;
3852 #ifdef FEAT_ARABIC
3853     char_u	*j;
3854 #endif
3855 
3856     if (*p < 0x80)		/* be quick for ASCII */
3857 	return 0;
3858 
3859     /* Skip backwards over trailing bytes: 10xx.xxxx
3860      * Skip backwards again if on a composing char. */
3861     for (q = p; ; --q)
3862     {
3863 	/* Move s to the last byte of this char. */
3864 	for (s = q; (s[1] & 0xc0) == 0x80; ++s)
3865 	    ;
3866 	/* Move q to the first byte of this char. */
3867 	while (q > base && (*q & 0xc0) == 0x80)
3868 	    --q;
3869 	/* Check for illegal sequence. Do allow an illegal byte after where we
3870 	 * started. */
3871 	len = utf8len_tab[*q];
3872 	if (len != (int)(s - q + 1) && len != (int)(p - q + 1))
3873 	    return 0;
3874 
3875 	if (q <= base)
3876 	    break;
3877 
3878 	c = utf_ptr2char(q);
3879 	if (utf_iscomposing(c))
3880 	    continue;
3881 
3882 #ifdef FEAT_ARABIC
3883 	if (arabic_maycombine(c))
3884 	{
3885 	    /* Advance to get a sneak-peak at the next char */
3886 	    j = q;
3887 	    --j;
3888 	    /* Move j to the first byte of this char. */
3889 	    while (j > base && (*j & 0xc0) == 0x80)
3890 		--j;
3891 	    if (arabic_combine(utf_ptr2char(j), c))
3892 		continue;
3893 	}
3894 #endif
3895 	break;
3896     }
3897 
3898     return (int)(p - q);
3899 }
3900 
3901 /*
3902  * Copy a character from "*fp" to "*tp" and advance the pointers.
3903  */
3904     void
3905 mb_copy_char(char_u **fp, char_u **tp)
3906 {
3907     int	    l = (*mb_ptr2len)(*fp);
3908 
3909     mch_memmove(*tp, *fp, (size_t)l);
3910     *tp += l;
3911     *fp += l;
3912 }
3913 
3914 /*
3915  * Return the offset from "p" to the first byte of a character.  When "p" is
3916  * at the start of a character 0 is returned, otherwise the offset to the next
3917  * character.  Can start anywhere in a stream of bytes.
3918  */
3919     int
3920 mb_off_next(char_u *base, char_u *p)
3921 {
3922     int		i;
3923     int		j;
3924 
3925     if (enc_utf8)
3926     {
3927 	if (*p < 0x80)		/* be quick for ASCII */
3928 	    return 0;
3929 
3930 	/* Find the next character that isn't 10xx.xxxx */
3931 	for (i = 0; (p[i] & 0xc0) == 0x80; ++i)
3932 	    ;
3933 	if (i > 0)
3934 	{
3935 	    /* Check for illegal sequence. */
3936 	    for (j = 0; p - j > base; ++j)
3937 		if ((p[-j] & 0xc0) != 0x80)
3938 		    break;
3939 	    if (utf8len_tab[p[-j]] != i + j)
3940 		return 0;
3941 	}
3942 	return i;
3943     }
3944 
3945     /* Only need to check if we're on a trail byte, it doesn't matter if we
3946      * want the offset to the next or current character. */
3947     return (*mb_head_off)(base, p);
3948 }
3949 
3950 /*
3951  * Return the offset from "p" to the last byte of the character it points
3952  * into.  Can start anywhere in a stream of bytes.
3953  */
3954     int
3955 mb_tail_off(char_u *base, char_u *p)
3956 {
3957     int		i;
3958     int		j;
3959 
3960     if (*p == NUL)
3961 	return 0;
3962 
3963     if (enc_utf8)
3964     {
3965 	/* Find the last character that is 10xx.xxxx */
3966 	for (i = 0; (p[i + 1] & 0xc0) == 0x80; ++i)
3967 	    ;
3968 	/* Check for illegal sequence. */
3969 	for (j = 0; p - j > base; ++j)
3970 	    if ((p[-j] & 0xc0) != 0x80)
3971 		break;
3972 	if (utf8len_tab[p[-j]] != i + j + 1)
3973 	    return 0;
3974 	return i;
3975     }
3976 
3977     /* It can't be the first byte if a double-byte when not using DBCS, at the
3978      * end of the string or the byte can't start a double-byte. */
3979     if (enc_dbcs == 0 || p[1] == NUL || MB_BYTE2LEN(*p) == 1)
3980 	return 0;
3981 
3982     /* Return 1 when on the lead byte, 0 when on the tail byte. */
3983     return 1 - dbcs_head_off(base, p);
3984 }
3985 
3986 /*
3987  * Find the next illegal byte sequence.
3988  */
3989     void
3990 utf_find_illegal(void)
3991 {
3992     pos_T	pos = curwin->w_cursor;
3993     char_u	*p;
3994     int		len;
3995     vimconv_T	vimconv;
3996     char_u	*tofree = NULL;
3997 
3998     vimconv.vc_type = CONV_NONE;
3999     if (enc_utf8 && (enc_canon_props(curbuf->b_p_fenc) & ENC_8BIT))
4000     {
4001 	/* 'encoding' is "utf-8" but we are editing a 8-bit encoded file,
4002 	 * possibly a utf-8 file with illegal bytes.  Setup for conversion
4003 	 * from utf-8 to 'fileencoding'. */
4004 	convert_setup(&vimconv, p_enc, curbuf->b_p_fenc);
4005     }
4006 
4007 #ifdef FEAT_VIRTUALEDIT
4008     curwin->w_cursor.coladd = 0;
4009 #endif
4010     for (;;)
4011     {
4012 	p = ml_get_cursor();
4013 	if (vimconv.vc_type != CONV_NONE)
4014 	{
4015 	    vim_free(tofree);
4016 	    tofree = string_convert(&vimconv, p, NULL);
4017 	    if (tofree == NULL)
4018 		break;
4019 	    p = tofree;
4020 	}
4021 
4022 	while (*p != NUL)
4023 	{
4024 	    /* Illegal means that there are not enough trail bytes (checked by
4025 	     * utf_ptr2len()) or too many of them (overlong sequence). */
4026 	    len = utf_ptr2len(p);
4027 	    if (*p >= 0x80 && (len == 1
4028 				     || utf_char2len(utf_ptr2char(p)) != len))
4029 	    {
4030 		if (vimconv.vc_type == CONV_NONE)
4031 		    curwin->w_cursor.col += (colnr_T)(p - ml_get_cursor());
4032 		else
4033 		{
4034 		    int	    l;
4035 
4036 		    len = (int)(p - tofree);
4037 		    for (p = ml_get_cursor(); *p != NUL && len-- > 0; p += l)
4038 		    {
4039 			l = utf_ptr2len(p);
4040 			curwin->w_cursor.col += l;
4041 		    }
4042 		}
4043 		goto theend;
4044 	    }
4045 	    p += len;
4046 	}
4047 	if (curwin->w_cursor.lnum == curbuf->b_ml.ml_line_count)
4048 	    break;
4049 	++curwin->w_cursor.lnum;
4050 	curwin->w_cursor.col = 0;
4051     }
4052 
4053     /* didn't find it: don't move and beep */
4054     curwin->w_cursor = pos;
4055     beep_flush();
4056 
4057 theend:
4058     vim_free(tofree);
4059     convert_setup(&vimconv, NULL, NULL);
4060 }
4061 
4062 #if defined(FEAT_GUI_GTK) || defined(PROTO)
4063 /*
4064  * Return TRUE if string "s" is a valid utf-8 string.
4065  * When "end" is NULL stop at the first NUL.
4066  * When "end" is positive stop there.
4067  */
4068     int
4069 utf_valid_string(char_u *s, char_u *end)
4070 {
4071     int		l;
4072     char_u	*p = s;
4073 
4074     while (end == NULL ? *p != NUL : p < end)
4075     {
4076 	l = utf8len_tab_zero[*p];
4077 	if (l == 0)
4078 	    return FALSE;	/* invalid lead byte */
4079 	if (end != NULL && p + l > end)
4080 	    return FALSE;	/* incomplete byte sequence */
4081 	++p;
4082 	while (--l > 0)
4083 	    if ((*p++ & 0xc0) != 0x80)
4084 		return FALSE;	/* invalid trail byte */
4085     }
4086     return TRUE;
4087 }
4088 #endif
4089 
4090 #if defined(FEAT_GUI) || defined(PROTO)
4091 /*
4092  * Special version of mb_tail_off() for use in ScreenLines[].
4093  */
4094     int
4095 dbcs_screen_tail_off(char_u *base, char_u *p)
4096 {
4097     /* It can't be the first byte if a double-byte when not using DBCS, at the
4098      * end of the string or the byte can't start a double-byte.
4099      * For euc-jp an 0x8e byte always means we have a lead byte in the current
4100      * cell. */
4101     if (*p == NUL || p[1] == NUL
4102 	    || (enc_dbcs == DBCS_JPNU && *p == 0x8e)
4103 	    || MB_BYTE2LEN(*p) == 1)
4104 	return 0;
4105 
4106     /* Return 1 when on the lead byte, 0 when on the tail byte. */
4107     return 1 - dbcs_screen_head_off(base, p);
4108 }
4109 #endif
4110 
4111 /*
4112  * If the cursor moves on an trail byte, set the cursor on the lead byte.
4113  * Thus it moves left if necessary.
4114  * Return TRUE when the cursor was adjusted.
4115  */
4116     void
4117 mb_adjust_cursor(void)
4118 {
4119     mb_adjustpos(curbuf, &curwin->w_cursor);
4120 }
4121 
4122 /*
4123  * Adjust position "*lp" to point to the first byte of a multi-byte character.
4124  * If it points to a tail byte it's moved backwards to the head byte.
4125  */
4126     void
4127 mb_adjustpos(buf_T *buf, pos_T *lp)
4128 {
4129     char_u	*p;
4130 
4131     if (lp->col > 0
4132 #ifdef FEAT_VIRTUALEDIT
4133 	    || lp->coladd > 1
4134 #endif
4135 	    )
4136     {
4137 	p = ml_get_buf(buf, lp->lnum, FALSE);
4138 	if (*p == NUL || (int)STRLEN(p) < lp->col)
4139 	    lp->col = 0;
4140 	else
4141 	    lp->col -= (*mb_head_off)(p, p + lp->col);
4142 #ifdef FEAT_VIRTUALEDIT
4143 	/* Reset "coladd" when the cursor would be on the right half of a
4144 	 * double-wide character. */
4145 	if (lp->coladd == 1
4146 		&& p[lp->col] != TAB
4147 		&& vim_isprintc((*mb_ptr2char)(p + lp->col))
4148 		&& ptr2cells(p + lp->col) > 1)
4149 	    lp->coladd = 0;
4150 #endif
4151     }
4152 }
4153 
4154 /*
4155  * Return a pointer to the character before "*p", if there is one.
4156  */
4157     char_u *
4158 mb_prevptr(
4159     char_u *line,	/* start of the string */
4160     char_u *p)
4161 {
4162     if (p > line)
4163 	MB_PTR_BACK(line, p);
4164     return p;
4165 }
4166 
4167 /*
4168  * Return the character length of "str".  Each multi-byte character (with
4169  * following composing characters) counts as one.
4170  */
4171     int
4172 mb_charlen(char_u *str)
4173 {
4174     char_u	*p = str;
4175     int		count;
4176 
4177     if (p == NULL)
4178 	return 0;
4179 
4180     for (count = 0; *p != NUL; count++)
4181 	p += (*mb_ptr2len)(p);
4182 
4183     return count;
4184 }
4185 
4186 #if defined(FEAT_SPELL) || defined(PROTO)
4187 /*
4188  * Like mb_charlen() but for a string with specified length.
4189  */
4190     int
4191 mb_charlen_len(char_u *str, int len)
4192 {
4193     char_u	*p = str;
4194     int		count;
4195 
4196     for (count = 0; *p != NUL && p < str + len; count++)
4197 	p += (*mb_ptr2len)(p);
4198 
4199     return count;
4200 }
4201 #endif
4202 
4203 /*
4204  * Try to un-escape a multi-byte character.
4205  * Used for the "to" and "from" part of a mapping.
4206  * Return the un-escaped string if it is a multi-byte character, and advance
4207  * "pp" to just after the bytes that formed it.
4208  * Return NULL if no multi-byte char was found.
4209  */
4210     char_u *
4211 mb_unescape(char_u **pp)
4212 {
4213     static char_u	buf[6];
4214     int			n;
4215     int			m = 0;
4216     char_u		*str = *pp;
4217 
4218     /* Must translate K_SPECIAL KS_SPECIAL KE_FILLER to K_SPECIAL and CSI
4219      * KS_EXTRA KE_CSI to CSI.
4220      * Maximum length of a utf-8 character is 4 bytes. */
4221     for (n = 0; str[n] != NUL && m < 4; ++n)
4222     {
4223 	if (str[n] == K_SPECIAL
4224 		&& str[n + 1] == KS_SPECIAL
4225 		&& str[n + 2] == KE_FILLER)
4226 	{
4227 	    buf[m++] = K_SPECIAL;
4228 	    n += 2;
4229 	}
4230 	else if ((str[n] == K_SPECIAL
4231 # ifdef FEAT_GUI
4232 		    || str[n] == CSI
4233 # endif
4234 		 )
4235 		&& str[n + 1] == KS_EXTRA
4236 		&& str[n + 2] == (int)KE_CSI)
4237 	{
4238 	    buf[m++] = CSI;
4239 	    n += 2;
4240 	}
4241 	else if (str[n] == K_SPECIAL
4242 # ifdef FEAT_GUI
4243 		|| str[n] == CSI
4244 # endif
4245 		)
4246 	    break;		/* a special key can't be a multibyte char */
4247 	else
4248 	    buf[m++] = str[n];
4249 	buf[m] = NUL;
4250 
4251 	/* Return a multi-byte character if it's found.  An illegal sequence
4252 	 * will result in a 1 here. */
4253 	if ((*mb_ptr2len)(buf) > 1)
4254 	{
4255 	    *pp = str + n + 1;
4256 	    return buf;
4257 	}
4258 
4259 	/* Bail out quickly for ASCII. */
4260 	if (buf[0] < 128)
4261 	    break;
4262     }
4263     return NULL;
4264 }
4265 
4266 /*
4267  * Return TRUE if the character at "row"/"col" on the screen is the left side
4268  * of a double-width character.
4269  * Caller must make sure "row" and "col" are not invalid!
4270  */
4271     int
4272 mb_lefthalve(int row, int col)
4273 {
4274 #ifdef FEAT_HANGULIN
4275     if (composing_hangul)
4276 	return TRUE;
4277 #endif
4278     return (*mb_off2cells)(LineOffset[row] + col,
4279 					LineOffset[row] + screen_Columns) > 1;
4280 }
4281 
4282 /*
4283  * Correct a position on the screen, if it's the right half of a double-wide
4284  * char move it to the left half.  Returns the corrected column.
4285  */
4286     int
4287 mb_fix_col(int col, int row)
4288 {
4289     col = check_col(col);
4290     row = check_row(row);
4291     if (has_mbyte && ScreenLines != NULL && col > 0
4292 	    && ((enc_dbcs
4293 		    && ScreenLines[LineOffset[row] + col] != NUL
4294 		    && dbcs_screen_head_off(ScreenLines + LineOffset[row],
4295 					 ScreenLines + LineOffset[row] + col))
4296 		|| (enc_utf8 && ScreenLines[LineOffset[row] + col] == 0)))
4297 	return col - 1;
4298     return col;
4299 }
4300 #endif
4301 
4302 #if defined(FEAT_MBYTE) || defined(FEAT_POSTSCRIPT) || defined(PROTO)
4303 static int enc_alias_search(char_u *name);
4304 
4305 /*
4306  * Skip the Vim specific head of a 'encoding' name.
4307  */
4308     char_u *
4309 enc_skip(char_u *p)
4310 {
4311     if (STRNCMP(p, "2byte-", 6) == 0)
4312 	return p + 6;
4313     if (STRNCMP(p, "8bit-", 5) == 0)
4314 	return p + 5;
4315     return p;
4316 }
4317 
4318 /*
4319  * Find the canonical name for encoding "enc".
4320  * When the name isn't recognized, returns "enc" itself, but with all lower
4321  * case characters and '_' replaced with '-'.
4322  * Returns an allocated string.  NULL for out-of-memory.
4323  */
4324     char_u *
4325 enc_canonize(char_u *enc)
4326 {
4327     char_u	*r;
4328     char_u	*p, *s;
4329     int		i;
4330 
4331 # ifdef FEAT_MBYTE
4332     if (STRCMP(enc, "default") == 0)
4333     {
4334 	/* Use the default encoding as it's found by set_init_1(). */
4335 	r = get_encoding_default();
4336 	if (r == NULL)
4337 	    r = (char_u *)"latin1";
4338 	return vim_strsave(r);
4339     }
4340 # endif
4341 
4342     /* copy "enc" to allocated memory, with room for two '-' */
4343     r = alloc((unsigned)(STRLEN(enc) + 3));
4344     if (r != NULL)
4345     {
4346 	/* Make it all lower case and replace '_' with '-'. */
4347 	p = r;
4348 	for (s = enc; *s != NUL; ++s)
4349 	{
4350 	    if (*s == '_')
4351 		*p++ = '-';
4352 	    else
4353 		*p++ = TOLOWER_ASC(*s);
4354 	}
4355 	*p = NUL;
4356 
4357 	/* Skip "2byte-" and "8bit-". */
4358 	p = enc_skip(r);
4359 
4360 	/* Change "microsoft-cp" to "cp".  Used in some spell files. */
4361 	if (STRNCMP(p, "microsoft-cp", 12) == 0)
4362 	    STRMOVE(p, p + 10);
4363 
4364 	/* "iso8859" -> "iso-8859" */
4365 	if (STRNCMP(p, "iso8859", 7) == 0)
4366 	{
4367 	    STRMOVE(p + 4, p + 3);
4368 	    p[3] = '-';
4369 	}
4370 
4371 	/* "iso-8859n" -> "iso-8859-n" */
4372 	if (STRNCMP(p, "iso-8859", 8) == 0 && p[8] != '-')
4373 	{
4374 	    STRMOVE(p + 9, p + 8);
4375 	    p[8] = '-';
4376 	}
4377 
4378 	/* "latin-N" -> "latinN" */
4379 	if (STRNCMP(p, "latin-", 6) == 0)
4380 	    STRMOVE(p + 5, p + 6);
4381 
4382 	if (enc_canon_search(p) >= 0)
4383 	{
4384 	    /* canonical name can be used unmodified */
4385 	    if (p != r)
4386 		STRMOVE(r, p);
4387 	}
4388 	else if ((i = enc_alias_search(p)) >= 0)
4389 	{
4390 	    /* alias recognized, get canonical name */
4391 	    vim_free(r);
4392 	    r = vim_strsave((char_u *)enc_canon_table[i].name);
4393 	}
4394     }
4395     return r;
4396 }
4397 
4398 /*
4399  * Search for an encoding alias of "name".
4400  * Returns -1 when not found.
4401  */
4402     static int
4403 enc_alias_search(char_u *name)
4404 {
4405     int		i;
4406 
4407     for (i = 0; enc_alias_table[i].name != NULL; ++i)
4408 	if (STRCMP(name, enc_alias_table[i].name) == 0)
4409 	    return enc_alias_table[i].canon;
4410     return -1;
4411 }
4412 #endif
4413 
4414 #if defined(FEAT_MBYTE) || defined(PROTO)
4415 
4416 # ifdef HAVE_LANGINFO_H
4417 #  include <langinfo.h>
4418 # endif
4419 
4420 # ifndef FEAT_GUI_W32
4421 /*
4422  * Get the canonicalized encoding from the specified locale string "locale"
4423  * or from the environment variables LC_ALL, LC_CTYPE and LANG.
4424  * Returns an allocated string when successful, NULL when not.
4425  */
4426     char_u *
4427 enc_locale_env(char *locale)
4428 {
4429     char	*s = locale;
4430     char	*p;
4431     int		i;
4432     char	buf[50];
4433 
4434     if (s == NULL || *s == NUL)
4435 	if ((s = getenv("LC_ALL")) == NULL || *s == NUL)
4436 	    if ((s = getenv("LC_CTYPE")) == NULL || *s == NUL)
4437 		s = getenv("LANG");
4438 
4439     if (s == NULL || *s == NUL)
4440 	return NULL;
4441 
4442     /* The most generic locale format is:
4443      * language[_territory][.codeset][@modifier][+special][,[sponsor][_revision]]
4444      * If there is a '.' remove the part before it.
4445      * if there is something after the codeset, remove it.
4446      * Make the name lowercase and replace '_' with '-'.
4447      * Exception: "ja_JP.EUC" == "euc-jp", "zh_CN.EUC" = "euc-cn",
4448      * "ko_KR.EUC" == "euc-kr"
4449      */
4450     if ((p = (char *)vim_strchr((char_u *)s, '.')) != NULL)
4451     {
4452 	if (p > s + 2 && STRNICMP(p + 1, "EUC", 3) == 0
4453 			&& !isalnum((int)p[4]) && p[4] != '-' && p[-3] == '_')
4454 	{
4455 	    /* copy "XY.EUC" to "euc-XY" to buf[10] */
4456 	    STRCPY(buf + 10, "euc-");
4457 	    buf[14] = p[-2];
4458 	    buf[15] = p[-1];
4459 	    buf[16] = 0;
4460 	    s = buf + 10;
4461 	}
4462 	else
4463 	    s = p + 1;
4464     }
4465     for (i = 0; i < (int)sizeof(buf) - 1 && s[i] != NUL; ++i)
4466     {
4467 	if (s[i] == '_' || s[i] == '-')
4468 	    buf[i] = '-';
4469 	else if (isalnum((int)s[i]))
4470 	    buf[i] = TOLOWER_ASC(s[i]);
4471 	else
4472 	    break;
4473     }
4474     buf[i] = NUL;
4475 
4476     return enc_canonize((char_u *)buf);
4477 }
4478 # endif
4479 
4480 /*
4481  * Get the canonicalized encoding of the current locale.
4482  * Returns an allocated string when successful, NULL when not.
4483  */
4484     char_u *
4485 enc_locale(void)
4486 {
4487 # ifdef WIN3264
4488     char	buf[50];
4489     long	acp = GetACP();
4490 
4491     if (acp == 1200)
4492 	STRCPY(buf, "ucs-2le");
4493     else if (acp == 1252)	    /* cp1252 is used as latin1 */
4494 	STRCPY(buf, "latin1");
4495     else
4496 	sprintf(buf, "cp%ld", acp);
4497 
4498     return enc_canonize((char_u *)buf);
4499 # else
4500     char	*s;
4501 
4502 #  ifdef HAVE_NL_LANGINFO_CODESET
4503     if ((s = nl_langinfo(CODESET)) == NULL || *s == NUL)
4504 #  endif
4505 #  if defined(HAVE_LOCALE_H) || defined(X_LOCALE)
4506 	if ((s = setlocale(LC_CTYPE, NULL)) == NULL || *s == NUL)
4507 #  endif
4508 	    s = NULL;
4509 
4510     return enc_locale_env(s);
4511 # endif
4512 }
4513 
4514 # if defined(WIN3264) || defined(PROTO) || defined(FEAT_CYGWIN_WIN32_CLIPBOARD)
4515 /*
4516  * Convert an encoding name to an MS-Windows codepage.
4517  * Returns zero if no codepage can be figured out.
4518  */
4519     int
4520 encname2codepage(char_u *name)
4521 {
4522     int		cp;
4523     char_u	*p = name;
4524     int		idx;
4525 
4526     if (STRNCMP(p, "8bit-", 5) == 0)
4527 	p += 5;
4528     else if (STRNCMP(p_enc, "2byte-", 6) == 0)
4529 	p += 6;
4530 
4531     if (p[0] == 'c' && p[1] == 'p')
4532 	cp = atoi((char *)p + 2);
4533     else if ((idx = enc_canon_search(p)) >= 0)
4534 	cp = enc_canon_table[idx].codepage;
4535     else
4536 	return 0;
4537     if (IsValidCodePage(cp))
4538 	return cp;
4539     return 0;
4540 }
4541 # endif
4542 
4543 # if defined(USE_ICONV) || defined(PROTO)
4544 
4545 static char_u *iconv_string(vimconv_T *vcp, char_u *str, int slen, int *unconvlenp, int *resultlenp);
4546 
4547 /*
4548  * Call iconv_open() with a check if iconv() works properly (there are broken
4549  * versions).
4550  * Returns (void *)-1 if failed.
4551  * (should return iconv_t, but that causes problems with prototypes).
4552  */
4553     void *
4554 my_iconv_open(char_u *to, char_u *from)
4555 {
4556     iconv_t	fd;
4557 #define ICONV_TESTLEN 400
4558     char_u	tobuf[ICONV_TESTLEN];
4559     char	*p;
4560     size_t	tolen;
4561     static int	iconv_ok = -1;
4562 
4563     if (iconv_ok == FALSE)
4564 	return (void *)-1;	/* detected a broken iconv() previously */
4565 
4566 #ifdef DYNAMIC_ICONV
4567     /* Check if the iconv.dll can be found. */
4568     if (!iconv_enabled(TRUE))
4569 	return (void *)-1;
4570 #endif
4571 
4572     fd = iconv_open((char *)enc_skip(to), (char *)enc_skip(from));
4573 
4574     if (fd != (iconv_t)-1 && iconv_ok == -1)
4575     {
4576 	/*
4577 	 * Do a dummy iconv() call to check if it actually works.  There is a
4578 	 * version of iconv() on Linux that is broken.  We can't ignore it,
4579 	 * because it's wide-spread.  The symptoms are that after outputting
4580 	 * the initial shift state the "to" pointer is NULL and conversion
4581 	 * stops for no apparent reason after about 8160 characters.
4582 	 */
4583 	p = (char *)tobuf;
4584 	tolen = ICONV_TESTLEN;
4585 	(void)iconv(fd, NULL, NULL, &p, &tolen);
4586 	if (p == NULL)
4587 	{
4588 	    iconv_ok = FALSE;
4589 	    iconv_close(fd);
4590 	    fd = (iconv_t)-1;
4591 	}
4592 	else
4593 	    iconv_ok = TRUE;
4594     }
4595 
4596     return (void *)fd;
4597 }
4598 
4599 /*
4600  * Convert the string "str[slen]" with iconv().
4601  * If "unconvlenp" is not NULL handle the string ending in an incomplete
4602  * sequence and set "*unconvlenp" to the length of it.
4603  * Returns the converted string in allocated memory.  NULL for an error.
4604  * If resultlenp is not NULL, sets it to the result length in bytes.
4605  */
4606     static char_u *
4607 iconv_string(
4608     vimconv_T	*vcp,
4609     char_u	*str,
4610     int		slen,
4611     int		*unconvlenp,
4612     int		*resultlenp)
4613 {
4614     const char	*from;
4615     size_t	fromlen;
4616     char	*to;
4617     size_t	tolen;
4618     size_t	len = 0;
4619     size_t	done = 0;
4620     char_u	*result = NULL;
4621     char_u	*p;
4622     int		l;
4623 
4624     from = (char *)str;
4625     fromlen = slen;
4626     for (;;)
4627     {
4628 	if (len == 0 || ICONV_ERRNO == ICONV_E2BIG)
4629 	{
4630 	    /* Allocate enough room for most conversions.  When re-allocating
4631 	     * increase the buffer size. */
4632 	    len = len + fromlen * 2 + 40;
4633 	    p = alloc((unsigned)len);
4634 	    if (p != NULL && done > 0)
4635 		mch_memmove(p, result, done);
4636 	    vim_free(result);
4637 	    result = p;
4638 	    if (result == NULL)	/* out of memory */
4639 		break;
4640 	}
4641 
4642 	to = (char *)result + done;
4643 	tolen = len - done - 2;
4644 	/* Avoid a warning for systems with a wrong iconv() prototype by
4645 	 * casting the second argument to void *. */
4646 	if (iconv(vcp->vc_fd, (void *)&from, &fromlen, &to, &tolen)
4647 								!= (size_t)-1)
4648 	{
4649 	    /* Finished, append a NUL. */
4650 	    *to = NUL;
4651 	    break;
4652 	}
4653 
4654 	/* Check both ICONV_EINVAL and EINVAL, because the dynamically loaded
4655 	 * iconv library may use one of them. */
4656 	if (!vcp->vc_fail && unconvlenp != NULL
4657 		&& (ICONV_ERRNO == ICONV_EINVAL || ICONV_ERRNO == EINVAL))
4658 	{
4659 	    /* Handle an incomplete sequence at the end. */
4660 	    *to = NUL;
4661 	    *unconvlenp = (int)fromlen;
4662 	    break;
4663 	}
4664 
4665 	/* Check both ICONV_EILSEQ and EILSEQ, because the dynamically loaded
4666 	 * iconv library may use one of them. */
4667 	else if (!vcp->vc_fail
4668 		&& (ICONV_ERRNO == ICONV_EILSEQ || ICONV_ERRNO == EILSEQ
4669 		    || ICONV_ERRNO == ICONV_EINVAL || ICONV_ERRNO == EINVAL))
4670 	{
4671 	    /* Can't convert: insert a '?' and skip a character.  This assumes
4672 	     * conversion from 'encoding' to something else.  In other
4673 	     * situations we don't know what to skip anyway. */
4674 	    *to++ = '?';
4675 	    if ((*mb_ptr2cells)((char_u *)from) > 1)
4676 		*to++ = '?';
4677 	    if (enc_utf8)
4678 		l = utfc_ptr2len_len((char_u *)from, (int)fromlen);
4679 	    else
4680 	    {
4681 		l = (*mb_ptr2len)((char_u *)from);
4682 		if (l > (int)fromlen)
4683 		    l = (int)fromlen;
4684 	    }
4685 	    from += l;
4686 	    fromlen -= l;
4687 	}
4688 	else if (ICONV_ERRNO != ICONV_E2BIG)
4689 	{
4690 	    /* conversion failed */
4691 	    VIM_CLEAR(result);
4692 	    break;
4693 	}
4694 	/* Not enough room or skipping illegal sequence. */
4695 	done = to - (char *)result;
4696     }
4697 
4698     if (resultlenp != NULL && result != NULL)
4699 	*resultlenp = (int)(to - (char *)result);
4700     return result;
4701 }
4702 
4703 #  if defined(DYNAMIC_ICONV) || defined(PROTO)
4704 /*
4705  * Dynamically load the "iconv.dll" on Win32.
4706  */
4707 
4708 #   ifndef DYNAMIC_ICONV	    /* must be generating prototypes */
4709 #    define HINSTANCE int
4710 #   endif
4711 static HINSTANCE hIconvDLL = 0;
4712 static HINSTANCE hMsvcrtDLL = 0;
4713 
4714 #   ifndef DYNAMIC_ICONV_DLL
4715 #    define DYNAMIC_ICONV_DLL "iconv.dll"
4716 #    define DYNAMIC_ICONV_DLL_ALT1 "libiconv.dll"
4717 #    define DYNAMIC_ICONV_DLL_ALT2 "libiconv2.dll"
4718 #    define DYNAMIC_ICONV_DLL_ALT3 "libiconv-2.dll"
4719 #   endif
4720 #   ifndef DYNAMIC_MSVCRT_DLL
4721 #    define DYNAMIC_MSVCRT_DLL "msvcrt.dll"
4722 #   endif
4723 
4724 /*
4725  * Try opening the iconv.dll and return TRUE if iconv() can be used.
4726  */
4727     int
4728 iconv_enabled(int verbose)
4729 {
4730     if (hIconvDLL != 0 && hMsvcrtDLL != 0)
4731 	return TRUE;
4732 
4733     /* The iconv DLL file goes under different names, try them all.
4734      * Do the "2" version first, it's newer. */
4735 #ifdef DYNAMIC_ICONV_DLL_ALT2
4736     if (hIconvDLL == 0)
4737 	hIconvDLL = vimLoadLib(DYNAMIC_ICONV_DLL_ALT2);
4738 #endif
4739 #ifdef DYNAMIC_ICONV_DLL_ALT3
4740     if (hIconvDLL == 0)
4741 	hIconvDLL = vimLoadLib(DYNAMIC_ICONV_DLL_ALT3);
4742 #endif
4743     if (hIconvDLL == 0)
4744 	hIconvDLL = vimLoadLib(DYNAMIC_ICONV_DLL);
4745 #ifdef DYNAMIC_ICONV_DLL_ALT1
4746     if (hIconvDLL == 0)
4747 	hIconvDLL = vimLoadLib(DYNAMIC_ICONV_DLL_ALT1);
4748 #endif
4749 
4750     if (hIconvDLL != 0)
4751 	hMsvcrtDLL = vimLoadLib(DYNAMIC_MSVCRT_DLL);
4752     if (hIconvDLL == 0 || hMsvcrtDLL == 0)
4753     {
4754 	/* Only give the message when 'verbose' is set, otherwise it might be
4755 	 * done whenever a conversion is attempted. */
4756 	if (verbose && p_verbose > 0)
4757 	{
4758 	    verbose_enter();
4759 	    EMSG2(_(e_loadlib),
4760 		    hIconvDLL == 0 ? DYNAMIC_ICONV_DLL : DYNAMIC_MSVCRT_DLL);
4761 	    verbose_leave();
4762 	}
4763 	iconv_end();
4764 	return FALSE;
4765     }
4766 
4767     iconv	= (void *)GetProcAddress(hIconvDLL, "libiconv");
4768     iconv_open	= (void *)GetProcAddress(hIconvDLL, "libiconv_open");
4769     iconv_close	= (void *)GetProcAddress(hIconvDLL, "libiconv_close");
4770     iconvctl	= (void *)GetProcAddress(hIconvDLL, "libiconvctl");
4771     iconv_errno	= get_dll_import_func(hIconvDLL, "_errno");
4772     if (iconv_errno == NULL)
4773 	iconv_errno = (void *)GetProcAddress(hMsvcrtDLL, "_errno");
4774     if (iconv == NULL || iconv_open == NULL || iconv_close == NULL
4775 	    || iconvctl == NULL || iconv_errno == NULL)
4776     {
4777 	iconv_end();
4778 	if (verbose && p_verbose > 0)
4779 	{
4780 	    verbose_enter();
4781 	    EMSG2(_(e_loadfunc), "for libiconv");
4782 	    verbose_leave();
4783 	}
4784 	return FALSE;
4785     }
4786     return TRUE;
4787 }
4788 
4789     void
4790 iconv_end(void)
4791 {
4792     /* Don't use iconv() when inputting or outputting characters. */
4793     if (input_conv.vc_type == CONV_ICONV)
4794 	convert_setup(&input_conv, NULL, NULL);
4795     if (output_conv.vc_type == CONV_ICONV)
4796 	convert_setup(&output_conv, NULL, NULL);
4797 
4798     if (hIconvDLL != 0)
4799 	FreeLibrary(hIconvDLL);
4800     if (hMsvcrtDLL != 0)
4801 	FreeLibrary(hMsvcrtDLL);
4802     hIconvDLL = 0;
4803     hMsvcrtDLL = 0;
4804 }
4805 #  endif /* DYNAMIC_ICONV */
4806 # endif /* USE_ICONV */
4807 
4808 #endif /* FEAT_MBYTE */
4809 
4810 #ifdef FEAT_GUI
4811 # define USE_IMACTIVATEFUNC (!gui.in_use && *p_imaf != NUL)
4812 # define USE_IMSTATUSFUNC (!gui.in_use && *p_imsf != NUL)
4813 #else
4814 # define USE_IMACTIVATEFUNC (*p_imaf != NUL)
4815 # define USE_IMSTATUSFUNC (*p_imsf != NUL)
4816 #endif
4817 
4818 #if defined(FEAT_EVAL) && defined(FEAT_MBYTE) \
4819 	&& (defined(FEAT_XIM) || defined(IME_WITHOUT_XIM))
4820     static void
4821 call_imactivatefunc(int active)
4822 {
4823     typval_T argv[2];
4824 
4825     argv[0].v_type = VAR_NUMBER;
4826     argv[0].vval.v_number = active ? 1 : 0;
4827     argv[1].v_type = VAR_UNKNOWN;
4828     (void)call_func_retnr(p_imaf, 1, argv);
4829 }
4830 
4831     static int
4832 call_imstatusfunc(void)
4833 {
4834     int is_active;
4835 
4836     /* FIXME: Don't execute user function in unsafe situation. */
4837     if (exiting || is_autocmd_blocked())
4838 	return FALSE;
4839     /* FIXME: :py print 'xxx' is shown duplicate result.
4840      * Use silent to avoid it. */
4841     ++msg_silent;
4842     is_active = call_func_retnr(p_imsf, 0, NULL);
4843     --msg_silent;
4844     return (is_active > 0);
4845 }
4846 #endif
4847 
4848 #if defined(FEAT_XIM) || defined(PROTO)
4849 
4850 # if defined(FEAT_GUI_GTK) || defined(PROTO)
4851 static int xim_has_preediting INIT(= FALSE);  /* IM current status */
4852 
4853 /*
4854  * Set preedit_start_col to the current cursor position.
4855  */
4856     static void
4857 init_preedit_start_col(void)
4858 {
4859     if (State & CMDLINE)
4860 	preedit_start_col = cmdline_getvcol_cursor();
4861     else if (curwin != NULL && curwin->w_buffer != NULL)
4862 	getvcol(curwin, &curwin->w_cursor, &preedit_start_col, NULL, NULL);
4863     /* Prevent that preediting marks the buffer as changed. */
4864     xim_changed_while_preediting = curbuf->b_changed;
4865 }
4866 
4867 static int im_is_active	       = FALSE;	/* IM is enabled for current mode    */
4868 static int preedit_is_active   = FALSE;
4869 static int im_preedit_cursor   = 0;	/* cursor offset in characters       */
4870 static int im_preedit_trailing = 0;	/* number of characters after cursor */
4871 
4872 static unsigned long im_commit_handler_id  = 0;
4873 static unsigned int  im_activatekey_keyval = GDK_VoidSymbol;
4874 static unsigned int  im_activatekey_state  = 0;
4875 
4876 static GtkWidget *preedit_window = NULL;
4877 static GtkWidget *preedit_label = NULL;
4878 
4879 static void im_preedit_window_set_position(void);
4880 
4881     void
4882 im_set_active(int active)
4883 {
4884     int was_active;
4885 
4886     was_active = !!im_get_status();
4887     im_is_active = (active && !p_imdisable);
4888 
4889     if (im_is_active != was_active)
4890 	xim_reset();
4891 }
4892 
4893     void
4894 xim_set_focus(int focus)
4895 {
4896     if (xic != NULL)
4897     {
4898 	if (focus)
4899 	    gtk_im_context_focus_in(xic);
4900 	else
4901 	    gtk_im_context_focus_out(xic);
4902     }
4903 }
4904 
4905     void
4906 im_set_position(int row, int col)
4907 {
4908     if (xic != NULL)
4909     {
4910 	GdkRectangle area;
4911 
4912 	area.x = FILL_X(col);
4913 	area.y = FILL_Y(row);
4914 	area.width  = gui.char_width * (mb_lefthalve(row, col) ? 2 : 1);
4915 	area.height = gui.char_height;
4916 
4917 	gtk_im_context_set_cursor_location(xic, &area);
4918 
4919 	if (p_imst == IM_OVER_THE_SPOT)
4920 	    im_preedit_window_set_position();
4921     }
4922 }
4923 
4924 #  if 0 || defined(PROTO) /* apparently only used in gui_x11.c */
4925     void
4926 xim_set_preedit(void)
4927 {
4928     im_set_position(gui.row, gui.col);
4929 }
4930 #  endif
4931 
4932     static void
4933 im_add_to_input(char_u *str, int len)
4934 {
4935     /* Convert from 'termencoding' (always "utf-8") to 'encoding' */
4936     if (input_conv.vc_type != CONV_NONE)
4937     {
4938 	str = string_convert(&input_conv, str, &len);
4939 	g_return_if_fail(str != NULL);
4940     }
4941 
4942     add_to_input_buf_csi(str, len);
4943 
4944     if (input_conv.vc_type != CONV_NONE)
4945 	vim_free(str);
4946 
4947     if (p_mh) /* blank out the pointer if necessary */
4948 	gui_mch_mousehide(TRUE);
4949 }
4950 
4951      static void
4952 im_preedit_window_set_position(void)
4953 {
4954     int x, y, width, height;
4955     int screen_x, screen_y, screen_width, screen_height;
4956 
4957     if (preedit_window == NULL)
4958 	return;
4959 
4960     gui_gtk_get_screen_geom_of_win(gui.drawarea,
4961 			  &screen_x, &screen_y, &screen_width, &screen_height);
4962 #if GTK_CHECK_VERSION(3,0,0)
4963     gdk_window_get_origin(gtk_widget_get_window(gui.drawarea), &x, &y);
4964 #else
4965     gdk_window_get_origin(gui.drawarea->window, &x, &y);
4966 #endif
4967     gtk_window_get_size(GTK_WINDOW(preedit_window), &width, &height);
4968     x = x + FILL_X(gui.col);
4969     y = y + FILL_Y(gui.row);
4970     if (x + width > screen_x + screen_width)
4971 	x = screen_x + screen_width - width;
4972     if (y + height > screen_y + screen_height)
4973 	y = screen_y + screen_height - height;
4974     gtk_window_move(GTK_WINDOW(preedit_window), x, y);
4975 }
4976 
4977     static void
4978 im_preedit_window_open()
4979 {
4980     char *preedit_string;
4981 #if !GTK_CHECK_VERSION(3,16,0)
4982     char buf[8];
4983 #endif
4984     PangoAttrList *attr_list;
4985     PangoLayout *layout;
4986 #if GTK_CHECK_VERSION(3,0,0)
4987 # if !GTK_CHECK_VERSION(3,16,0)
4988     GdkRGBA color;
4989 # endif
4990 #else
4991     GdkColor color;
4992 #endif
4993     gint w, h;
4994 
4995     if (preedit_window == NULL)
4996     {
4997 	preedit_window = gtk_window_new(GTK_WINDOW_POPUP);
4998 	gtk_window_set_transient_for(GTK_WINDOW(preedit_window),
4999 						     GTK_WINDOW(gui.mainwin));
5000 	preedit_label = gtk_label_new("");
5001 	gtk_widget_set_name(preedit_label, "vim-gui-preedit-area");
5002 	gtk_container_add(GTK_CONTAINER(preedit_window), preedit_label);
5003     }
5004 
5005 #if GTK_CHECK_VERSION(3,16,0)
5006     {
5007 	GtkStyleContext * const context
5008 				  = gtk_widget_get_style_context(gui.drawarea);
5009 	GtkCssProvider * const provider = gtk_css_provider_new();
5010 	gchar		   *css = NULL;
5011 	const char * const fontname
5012 			   = pango_font_description_get_family(gui.norm_font);
5013 	gint	fontsize
5014 		= pango_font_description_get_size(gui.norm_font) / PANGO_SCALE;
5015 	gchar	*fontsize_propval = NULL;
5016 
5017 	if (!pango_font_description_get_size_is_absolute(gui.norm_font))
5018 	{
5019 	    /* fontsize was given in points.  Convert it into that in pixels
5020 	     * to use with CSS. */
5021 	    GdkScreen * const screen
5022 		  = gdk_window_get_screen(gtk_widget_get_window(gui.mainwin));
5023 	    const gdouble dpi = gdk_screen_get_resolution(screen);
5024 	    fontsize = dpi * fontsize / 72;
5025 	}
5026 	if (fontsize > 0)
5027 	    fontsize_propval = g_strdup_printf("%dpx", fontsize);
5028 	else
5029 	    fontsize_propval = g_strdup_printf("inherit");
5030 
5031 	css = g_strdup_printf(
5032 		"widget#vim-gui-preedit-area {\n"
5033 		"  font-family: %s,monospace;\n"
5034 		"  font-size: %s;\n"
5035 		"  color: #%.2lx%.2lx%.2lx;\n"
5036 		"  background-color: #%.2lx%.2lx%.2lx;\n"
5037 		"}\n",
5038 		fontname != NULL ? fontname : "inherit",
5039 		fontsize_propval,
5040 		(gui.norm_pixel >> 16) & 0xff,
5041 		(gui.norm_pixel >> 8) & 0xff,
5042 		gui.norm_pixel & 0xff,
5043 		(gui.back_pixel >> 16) & 0xff,
5044 		(gui.back_pixel >> 8) & 0xff,
5045 		gui.back_pixel & 0xff);
5046 
5047 	gtk_css_provider_load_from_data(provider, css, -1, NULL);
5048 	gtk_style_context_add_provider(context,
5049 				     GTK_STYLE_PROVIDER(provider), G_MAXUINT);
5050 
5051 	g_free(css);
5052 	g_free(fontsize_propval);
5053 	g_object_unref(provider);
5054     }
5055 #elif GTK_CHECK_VERSION(3,0,0)
5056     gtk_widget_override_font(preedit_label, gui.norm_font);
5057 
5058     vim_snprintf(buf, sizeof(buf), "#%06X", gui.norm_pixel);
5059     gdk_rgba_parse(&color, buf);
5060     gtk_widget_override_color(preedit_label, GTK_STATE_FLAG_NORMAL, &color);
5061 
5062     vim_snprintf(buf, sizeof(buf), "#%06X", gui.back_pixel);
5063     gdk_rgba_parse(&color, buf);
5064     gtk_widget_override_background_color(preedit_label, GTK_STATE_FLAG_NORMAL,
5065 								      &color);
5066 #else
5067     gtk_widget_modify_font(preedit_label, gui.norm_font);
5068 
5069     vim_snprintf(buf, sizeof(buf), "#%06X", (unsigned)gui.norm_pixel);
5070     gdk_color_parse(buf, &color);
5071     gtk_widget_modify_fg(preedit_label, GTK_STATE_NORMAL, &color);
5072 
5073     vim_snprintf(buf, sizeof(buf), "#%06X", (unsigned)gui.back_pixel);
5074     gdk_color_parse(buf, &color);
5075     gtk_widget_modify_bg(preedit_window, GTK_STATE_NORMAL, &color);
5076 #endif
5077 
5078     gtk_im_context_get_preedit_string(xic, &preedit_string, &attr_list, NULL);
5079 
5080     if (preedit_string[0] != NUL)
5081     {
5082 	gtk_label_set_text(GTK_LABEL(preedit_label), preedit_string);
5083 	gtk_label_set_attributes(GTK_LABEL(preedit_label), attr_list);
5084 
5085 	layout = gtk_label_get_layout(GTK_LABEL(preedit_label));
5086 	pango_layout_get_pixel_size(layout, &w, &h);
5087 	h = MAX(h, gui.char_height);
5088 	gtk_window_resize(GTK_WINDOW(preedit_window), w, h);
5089 
5090 	gtk_widget_show_all(preedit_window);
5091 
5092 	im_preedit_window_set_position();
5093     }
5094 
5095     g_free(preedit_string);
5096     pango_attr_list_unref(attr_list);
5097 }
5098 
5099     static void
5100 im_preedit_window_close()
5101 {
5102     if (preedit_window != NULL)
5103 	gtk_widget_hide(preedit_window);
5104 }
5105 
5106     static void
5107 im_show_preedit()
5108 {
5109     im_preedit_window_open();
5110 
5111     if (p_mh) /* blank out the pointer if necessary */
5112 	gui_mch_mousehide(TRUE);
5113 }
5114 
5115     static void
5116 im_delete_preedit(void)
5117 {
5118     char_u bskey[]  = {CSI, 'k', 'b'};
5119     char_u delkey[] = {CSI, 'k', 'D'};
5120 
5121     if (p_imst == IM_OVER_THE_SPOT)
5122     {
5123 	im_preedit_window_close();
5124 	return;
5125     }
5126 
5127     if (State & NORMAL
5128 #ifdef FEAT_TERMINAL
5129 	    && !term_use_loop()
5130 #endif
5131        )
5132     {
5133 	im_preedit_cursor = 0;
5134 	return;
5135     }
5136     for (; im_preedit_cursor > 0; --im_preedit_cursor)
5137 	add_to_input_buf(bskey, (int)sizeof(bskey));
5138 
5139     for (; im_preedit_trailing > 0; --im_preedit_trailing)
5140 	add_to_input_buf(delkey, (int)sizeof(delkey));
5141 }
5142 
5143 /*
5144  * Move the cursor left by "num_move_back" characters.
5145  * Note that ins_left() checks im_is_preediting() to avoid breaking undo for
5146  * these K_LEFT keys.
5147  */
5148     static void
5149 im_correct_cursor(int num_move_back)
5150 {
5151     char_u backkey[] = {CSI, 'k', 'l'};
5152 
5153     if (State & NORMAL)
5154 	return;
5155 #  ifdef FEAT_RIGHTLEFT
5156     if ((State & CMDLINE) == 0 && curwin != NULL && curwin->w_p_rl)
5157 	backkey[2] = 'r';
5158 #  endif
5159     for (; num_move_back > 0; --num_move_back)
5160 	add_to_input_buf(backkey, (int)sizeof(backkey));
5161 }
5162 
5163 static int xim_expected_char = NUL;
5164 static int xim_ignored_char = FALSE;
5165 
5166 /*
5167  * Update the mode and cursor while in an IM callback.
5168  */
5169     static void
5170 im_show_info(void)
5171 {
5172     int	    old_vgetc_busy;
5173 
5174     old_vgetc_busy = vgetc_busy;
5175     vgetc_busy = TRUE;
5176     showmode();
5177     vgetc_busy = old_vgetc_busy;
5178     if ((State & NORMAL) || (State & INSERT))
5179 	setcursor();
5180     out_flush();
5181 }
5182 
5183 /*
5184  * Callback invoked when the user finished preediting.
5185  * Put the final string into the input buffer.
5186  */
5187     static void
5188 im_commit_cb(GtkIMContext *context UNUSED,
5189 	     const gchar *str,
5190 	     gpointer data UNUSED)
5191 {
5192     int		slen = (int)STRLEN(str);
5193     int		add_to_input = TRUE;
5194     int		clen;
5195     int		len = slen;
5196     int		commit_with_preedit = TRUE;
5197     char_u	*im_str;
5198 
5199 #ifdef XIM_DEBUG
5200     xim_log("im_commit_cb(): %s\n", str);
5201 #endif
5202 
5203     if (p_imst == IM_ON_THE_SPOT)
5204     {
5205 	/* The imhangul module doesn't reset the preedit string before
5206 	 * committing.  Call im_delete_preedit() to work around that. */
5207 	im_delete_preedit();
5208 
5209 	/* Indicate that preediting has finished. */
5210 	if (preedit_start_col == MAXCOL)
5211 	{
5212 	    init_preedit_start_col();
5213 	    commit_with_preedit = FALSE;
5214 	}
5215 
5216 	/* The thing which setting "preedit_start_col" to MAXCOL means that
5217 	 * "preedit_start_col" will be set forcedly when calling
5218 	 * preedit_changed_cb() next time.
5219 	 * "preedit_start_col" should not reset with MAXCOL on this part. Vim
5220 	 * is simulating the preediting by using add_to_input_str(). when
5221 	 * preedit begin immediately before committed, the typebuf is not
5222 	 * flushed to screen, then it can't get correct "preedit_start_col".
5223 	 * Thus, it should calculate the cells by adding cells of the committed
5224 	 * string. */
5225 	if (input_conv.vc_type != CONV_NONE)
5226 	{
5227 	    im_str = string_convert(&input_conv, (char_u *)str, &len);
5228 	    g_return_if_fail(im_str != NULL);
5229 	}
5230 	else
5231 	    im_str = (char_u *)str;
5232 
5233 	clen = mb_string2cells(im_str, len);
5234 
5235 	if (input_conv.vc_type != CONV_NONE)
5236 	    vim_free(im_str);
5237 	preedit_start_col += clen;
5238     }
5239 
5240     /* Is this a single character that matches a keypad key that's just
5241      * been pressed?  If so, we don't want it to be entered as such - let
5242      * us carry on processing the raw keycode so that it may be used in
5243      * mappings as <kSomething>. */
5244     if (xim_expected_char != NUL)
5245     {
5246 	/* We're currently processing a keypad or other special key */
5247 	if (slen == 1 && str[0] == xim_expected_char)
5248 	{
5249 	    /* It's a match - don't do it here */
5250 	    xim_ignored_char = TRUE;
5251 	    add_to_input = FALSE;
5252 	}
5253 	else
5254 	{
5255 	    /* Not a match */
5256 	    xim_ignored_char = FALSE;
5257 	}
5258     }
5259 
5260     if (add_to_input)
5261 	im_add_to_input((char_u *)str, slen);
5262 
5263     if (p_imst == IM_ON_THE_SPOT)
5264     {
5265 	/* Inserting chars while "im_is_active" is set does not cause a
5266 	 * change of buffer.  When the chars are committed the buffer must be
5267 	 * marked as changed. */
5268 	if (!commit_with_preedit)
5269 	    preedit_start_col = MAXCOL;
5270 
5271 	/* This flag is used in changed() at next call. */
5272 	xim_changed_while_preediting = TRUE;
5273     }
5274 
5275     if (gtk_main_level() > 0)
5276 	gtk_main_quit();
5277 }
5278 
5279 /*
5280  * Callback invoked after start to the preedit.
5281  */
5282     static void
5283 im_preedit_start_cb(GtkIMContext *context UNUSED, gpointer data UNUSED)
5284 {
5285 #ifdef XIM_DEBUG
5286     xim_log("im_preedit_start_cb()\n");
5287 #endif
5288 
5289     im_is_active = TRUE;
5290     preedit_is_active = TRUE;
5291     gui_update_cursor(TRUE, FALSE);
5292     im_show_info();
5293 }
5294 
5295 /*
5296  * Callback invoked after end to the preedit.
5297  */
5298     static void
5299 im_preedit_end_cb(GtkIMContext *context UNUSED, gpointer data UNUSED)
5300 {
5301 #ifdef XIM_DEBUG
5302     xim_log("im_preedit_end_cb()\n");
5303 #endif
5304     im_delete_preedit();
5305 
5306     /* Indicate that preediting has finished */
5307     if (p_imst == IM_ON_THE_SPOT)
5308 	preedit_start_col = MAXCOL;
5309     xim_has_preediting = FALSE;
5310 
5311 #if 0
5312     /* Removal of this line suggested by Takuhiro Nishioka.  Fixes that IM was
5313      * switched off unintentionally.  We now use preedit_is_active (added by
5314      * SungHyun Nam). */
5315     im_is_active = FALSE;
5316 #endif
5317     preedit_is_active = FALSE;
5318     gui_update_cursor(TRUE, FALSE);
5319     im_show_info();
5320 }
5321 
5322 /*
5323  * Callback invoked after changes to the preedit string.  If the preedit
5324  * string was empty before, remember the preedit start column so we know
5325  * where to apply feedback attributes.  Delete the previous preedit string
5326  * if there was one, save the new preedit cursor offset, and put the new
5327  * string into the input buffer.
5328  *
5329  * TODO: The pragmatic "put into input buffer" approach used here has
5330  *       several fundamental problems:
5331  *
5332  * - The characters in the preedit string are subject to remapping.
5333  *   That's broken, only the finally committed string should be remapped.
5334  *
5335  * - There is a race condition involved:  The retrieved value for the
5336  *   current cursor position will be wrong if any unprocessed characters
5337  *   are still queued in the input buffer.
5338  *
5339  * - Due to the lack of synchronization between the file buffer in memory
5340  *   and any typed characters, it's practically impossible to implement the
5341  *   "retrieve_surrounding" and "delete_surrounding" signals reliably.  IM
5342  *   modules for languages such as Thai are likely to rely on this feature
5343  *   for proper operation.
5344  *
5345  * Conclusions:  I think support for preediting needs to be moved to the
5346  * core parts of Vim.  Ideally, until it has been committed, the preediting
5347  * string should only be displayed and not affect the buffer content at all.
5348  * The question how to deal with the synchronization issue still remains.
5349  * Circumventing the input buffer is probably not desirable.  Anyway, I think
5350  * implementing "retrieve_surrounding" is the only hard problem.
5351  *
5352  * One way to solve all of this in a clean manner would be to queue all key
5353  * press/release events "as is" in the input buffer, and apply the IM filtering
5354  * at the receiving end of the queue.  This, however, would have a rather large
5355  * impact on the code base.  If there is an easy way to force processing of all
5356  * remaining input from within the "retrieve_surrounding" signal handler, this
5357  * might not be necessary.  Gotta ask on vim-dev for opinions.
5358  */
5359     static void
5360 im_preedit_changed_cb(GtkIMContext *context, gpointer data UNUSED)
5361 {
5362     char    *preedit_string = NULL;
5363     int	    cursor_index    = 0;
5364     int	    num_move_back   = 0;
5365     char_u  *str;
5366     char_u  *p;
5367     int	    i;
5368 
5369     if (p_imst == IM_ON_THE_SPOT)
5370 	gtk_im_context_get_preedit_string(context,
5371 					  &preedit_string, NULL,
5372 					  &cursor_index);
5373     else
5374 	gtk_im_context_get_preedit_string(context,
5375 					  &preedit_string, NULL,
5376 					  NULL);
5377 
5378 #ifdef XIM_DEBUG
5379     xim_log("im_preedit_changed_cb(): %s\n", preedit_string);
5380 #endif
5381 
5382     g_return_if_fail(preedit_string != NULL); /* just in case */
5383 
5384     if (p_imst == IM_OVER_THE_SPOT)
5385     {
5386 	if (preedit_string[0] == NUL)
5387 	{
5388 	    xim_has_preediting = FALSE;
5389 	    im_delete_preedit();
5390 	}
5391 	else
5392 	{
5393 	    xim_has_preediting = TRUE;
5394 	    im_show_preedit();
5395 	}
5396     }
5397     else
5398     {
5399 	/* If preedit_start_col is MAXCOL set it to the current cursor position. */
5400 	if (preedit_start_col == MAXCOL && preedit_string[0] != '\0')
5401 	{
5402 	    xim_has_preediting = TRUE;
5403 
5404 	    /* Urgh, this breaks if the input buffer isn't empty now */
5405 	    init_preedit_start_col();
5406 	}
5407 	else if (cursor_index == 0 && preedit_string[0] == '\0')
5408 	{
5409 	    xim_has_preediting = FALSE;
5410 
5411 	    /* If at the start position (after typing backspace)
5412 	     * preedit_start_col must be reset. */
5413 	    preedit_start_col = MAXCOL;
5414 	}
5415 
5416 	im_delete_preedit();
5417 
5418 	/*
5419 	 * Compute the end of the preediting area: "preedit_end_col".
5420 	 * According to the documentation of gtk_im_context_get_preedit_string(),
5421 	 * the cursor_pos output argument returns the offset in bytes.  This is
5422 	 * unfortunately not true -- real life shows the offset is in characters,
5423 	 * and the GTK+ source code agrees with me.  Will file a bug later.
5424 	 */
5425 	if (preedit_start_col != MAXCOL)
5426 	    preedit_end_col = preedit_start_col;
5427 	str = (char_u *)preedit_string;
5428 	for (p = str, i = 0; *p != NUL; p += utf_byte2len(*p), ++i)
5429 	{
5430 	    int is_composing;
5431 
5432 	    is_composing = ((*p & 0x80) != 0 && utf_iscomposing(utf_ptr2char(p)));
5433 	    /*
5434 	     * These offsets are used as counters when generating <BS> and <Del>
5435 	     * to delete the preedit string.  So don't count composing characters
5436 	     * unless 'delcombine' is enabled.
5437 	     */
5438 	    if (!is_composing || p_deco)
5439 	    {
5440 		if (i < cursor_index)
5441 		    ++im_preedit_cursor;
5442 		else
5443 		    ++im_preedit_trailing;
5444 	    }
5445 	    if (!is_composing && i >= cursor_index)
5446 	    {
5447 		/* This is essentially the same as im_preedit_trailing, except
5448 		 * composing characters are not counted even if p_deco is set. */
5449 		++num_move_back;
5450 	    }
5451 	    if (preedit_start_col != MAXCOL)
5452 		preedit_end_col += utf_ptr2cells(p);
5453 	}
5454 
5455 	if (p > str)
5456 	{
5457 	    im_add_to_input(str, (int)(p - str));
5458 	    im_correct_cursor(num_move_back);
5459 	}
5460     }
5461 
5462     g_free(preedit_string);
5463 
5464     if (gtk_main_level() > 0)
5465 	gtk_main_quit();
5466 }
5467 
5468 /*
5469  * Translate the Pango attributes at iter to Vim highlighting attributes.
5470  * Ignore attributes not supported by Vim highlighting.  This shouldn't have
5471  * too much impact -- right now we handle even more attributes than necessary
5472  * for the IM modules I tested with.
5473  */
5474     static int
5475 translate_pango_attributes(PangoAttrIterator *iter)
5476 {
5477     PangoAttribute  *attr;
5478     int		    char_attr = HL_NORMAL;
5479 
5480     attr = pango_attr_iterator_get(iter, PANGO_ATTR_UNDERLINE);
5481     if (attr != NULL && ((PangoAttrInt *)attr)->value
5482 						 != (int)PANGO_UNDERLINE_NONE)
5483 	char_attr |= HL_UNDERLINE;
5484 
5485     attr = pango_attr_iterator_get(iter, PANGO_ATTR_WEIGHT);
5486     if (attr != NULL && ((PangoAttrInt *)attr)->value >= (int)PANGO_WEIGHT_BOLD)
5487 	char_attr |= HL_BOLD;
5488 
5489     attr = pango_attr_iterator_get(iter, PANGO_ATTR_STYLE);
5490     if (attr != NULL && ((PangoAttrInt *)attr)->value
5491 						   != (int)PANGO_STYLE_NORMAL)
5492 	char_attr |= HL_ITALIC;
5493 
5494     attr = pango_attr_iterator_get(iter, PANGO_ATTR_BACKGROUND);
5495     if (attr != NULL)
5496     {
5497 	const PangoColor *color = &((PangoAttrColor *)attr)->color;
5498 
5499 	/* Assume inverse if black background is requested */
5500 	if ((color->red | color->green | color->blue) == 0)
5501 	    char_attr |= HL_INVERSE;
5502     }
5503 
5504     return char_attr;
5505 }
5506 
5507 /*
5508  * Retrieve the highlighting attributes at column col in the preedit string.
5509  * Return -1 if not in preediting mode or if col is out of range.
5510  */
5511     int
5512 im_get_feedback_attr(int col)
5513 {
5514     char	    *preedit_string = NULL;
5515     PangoAttrList   *attr_list	    = NULL;
5516     int		    char_attr	    = -1;
5517 
5518     if (xic == NULL)
5519 	return char_attr;
5520 
5521     gtk_im_context_get_preedit_string(xic, &preedit_string, &attr_list, NULL);
5522 
5523     if (preedit_string != NULL && attr_list != NULL)
5524     {
5525 	int idx;
5526 
5527 	/* Get the byte index as used by PangoAttrIterator */
5528 	for (idx = 0; col > 0 && preedit_string[idx] != '\0'; --col)
5529 	    idx += utfc_ptr2len((char_u *)preedit_string + idx);
5530 
5531 	if (preedit_string[idx] != '\0')
5532 	{
5533 	    PangoAttrIterator	*iter;
5534 	    int			start, end;
5535 
5536 	    char_attr = HL_NORMAL;
5537 	    iter = pango_attr_list_get_iterator(attr_list);
5538 
5539 	    /* Extract all relevant attributes from the list. */
5540 	    do
5541 	    {
5542 		pango_attr_iterator_range(iter, &start, &end);
5543 
5544 		if (idx >= start && idx < end)
5545 		    char_attr |= translate_pango_attributes(iter);
5546 	    }
5547 	    while (pango_attr_iterator_next(iter));
5548 
5549 	    pango_attr_iterator_destroy(iter);
5550 	}
5551     }
5552 
5553     if (attr_list != NULL)
5554 	pango_attr_list_unref(attr_list);
5555     g_free(preedit_string);
5556 
5557     return char_attr;
5558 }
5559 
5560     void
5561 xim_init(void)
5562 {
5563 #ifdef XIM_DEBUG
5564     xim_log("xim_init()\n");
5565 #endif
5566 
5567     g_return_if_fail(gui.drawarea != NULL);
5568 #if GTK_CHECK_VERSION(3,0,0)
5569     g_return_if_fail(gtk_widget_get_window(gui.drawarea) != NULL);
5570 #else
5571     g_return_if_fail(gui.drawarea->window != NULL);
5572 #endif
5573 
5574     xic = gtk_im_multicontext_new();
5575     g_object_ref(xic);
5576 
5577     im_commit_handler_id = g_signal_connect(G_OBJECT(xic), "commit",
5578 					    G_CALLBACK(&im_commit_cb), NULL);
5579     g_signal_connect(G_OBJECT(xic), "preedit_changed",
5580 		     G_CALLBACK(&im_preedit_changed_cb), NULL);
5581     g_signal_connect(G_OBJECT(xic), "preedit_start",
5582 		     G_CALLBACK(&im_preedit_start_cb), NULL);
5583     g_signal_connect(G_OBJECT(xic), "preedit_end",
5584 		     G_CALLBACK(&im_preedit_end_cb), NULL);
5585 
5586 #if GTK_CHECK_VERSION(3,0,0)
5587     gtk_im_context_set_client_window(xic, gtk_widget_get_window(gui.drawarea));
5588 #else
5589     gtk_im_context_set_client_window(xic, gui.drawarea->window);
5590 #endif
5591 }
5592 
5593     void
5594 im_shutdown(void)
5595 {
5596 #ifdef XIM_DEBUG
5597     xim_log("im_shutdown()\n");
5598 #endif
5599 
5600     if (xic != NULL)
5601     {
5602 	gtk_im_context_focus_out(xic);
5603 	g_object_unref(xic);
5604 	xic = NULL;
5605     }
5606     im_is_active = FALSE;
5607     im_commit_handler_id = 0;
5608     if (p_imst == IM_ON_THE_SPOT)
5609 	preedit_start_col = MAXCOL;
5610     xim_has_preediting = FALSE;
5611 }
5612 
5613 /*
5614  * Convert the string argument to keyval and state for GdkEventKey.
5615  * If str is valid return TRUE, otherwise FALSE.
5616  *
5617  * See 'imactivatekey' for documentation of the format.
5618  */
5619     static int
5620 im_string_to_keyval(const char *str, unsigned int *keyval, unsigned int *state)
5621 {
5622     const char	    *mods_end;
5623     unsigned	    tmp_keyval;
5624     unsigned	    tmp_state = 0;
5625 
5626     mods_end = strrchr(str, '-');
5627     mods_end = (mods_end != NULL) ? mods_end + 1 : str;
5628 
5629     /* Parse modifier keys */
5630     while (str < mods_end)
5631 	switch (*str++)
5632 	{
5633 	    case '-':							break;
5634 	    case 'S': case 's': tmp_state |= (unsigned)GDK_SHIFT_MASK;	break;
5635 	    case 'L': case 'l': tmp_state |= (unsigned)GDK_LOCK_MASK;	break;
5636 	    case 'C': case 'c': tmp_state |= (unsigned)GDK_CONTROL_MASK;break;
5637 	    case '1':		tmp_state |= (unsigned)GDK_MOD1_MASK;	break;
5638 	    case '2':		tmp_state |= (unsigned)GDK_MOD2_MASK;	break;
5639 	    case '3':		tmp_state |= (unsigned)GDK_MOD3_MASK;	break;
5640 	    case '4':		tmp_state |= (unsigned)GDK_MOD4_MASK;	break;
5641 	    case '5':		tmp_state |= (unsigned)GDK_MOD5_MASK;	break;
5642 	    default:
5643 		return FALSE;
5644 	}
5645 
5646     tmp_keyval = gdk_keyval_from_name(str);
5647 
5648     if (tmp_keyval == 0 || tmp_keyval == GDK_VoidSymbol)
5649 	return FALSE;
5650 
5651     if (keyval != NULL)
5652 	*keyval = tmp_keyval;
5653     if (state != NULL)
5654 	*state = tmp_state;
5655 
5656     return TRUE;
5657 }
5658 
5659 /*
5660  * Return TRUE if p_imak is valid, otherwise FALSE.  As a special case, an
5661  * empty string is also regarded as valid.
5662  *
5663  * Note: The numerical key value of p_imak is cached if it was valid; thus
5664  * boldly assuming im_xim_isvalid_imactivate() will always be called whenever
5665  * 'imak' changes.  This is currently the case but not obvious -- should
5666  * probably rename the function for clarity.
5667  */
5668     int
5669 im_xim_isvalid_imactivate(void)
5670 {
5671     if (p_imak[0] == NUL)
5672     {
5673 	im_activatekey_keyval = GDK_VoidSymbol;
5674 	im_activatekey_state  = 0;
5675 	return TRUE;
5676     }
5677 
5678     return im_string_to_keyval((const char *)p_imak,
5679 			       &im_activatekey_keyval,
5680 			       &im_activatekey_state);
5681 }
5682 
5683     static void
5684 im_synthesize_keypress(unsigned int keyval, unsigned int state)
5685 {
5686     GdkEventKey *event;
5687 
5688     event = (GdkEventKey *)gdk_event_new(GDK_KEY_PRESS);
5689 #  if GTK_CHECK_VERSION(3,0,0)
5690     g_object_ref(gtk_widget_get_window(gui.drawarea));
5691 					/* unreffed by gdk_event_free() */
5692 #  else
5693     g_object_ref(gui.drawarea->window); /* unreffed by gdk_event_free() */
5694 #  endif
5695 #  if GTK_CHECK_VERSION(3,0,0)
5696     event->window = gtk_widget_get_window(gui.drawarea);
5697 #  else
5698     event->window = gui.drawarea->window;
5699 #  endif
5700     event->send_event = TRUE;
5701     event->time = GDK_CURRENT_TIME;
5702     event->state  = state;
5703     event->keyval = keyval;
5704     event->hardware_keycode = /* needed for XIM */
5705 	XKeysymToKeycode(GDK_WINDOW_XDISPLAY(event->window), (KeySym)keyval);
5706     event->length = 0;
5707     event->string = NULL;
5708 
5709     gtk_im_context_filter_keypress(xic, event);
5710 
5711     /* For consistency, also send the corresponding release event. */
5712     event->type = GDK_KEY_RELEASE;
5713     event->send_event = FALSE;
5714     gtk_im_context_filter_keypress(xic, event);
5715 
5716     gdk_event_free((GdkEvent *)event);
5717 }
5718 
5719     void
5720 xim_reset(void)
5721 {
5722 # ifdef FEAT_EVAL
5723     if (USE_IMACTIVATEFUNC)
5724 	call_imactivatefunc(im_is_active);
5725     else
5726 # endif
5727     if (xic != NULL)
5728     {
5729 	gtk_im_context_reset(xic);
5730 
5731 	if (p_imdisable)
5732 	    im_shutdown();
5733 	else
5734 	{
5735 	    xim_set_focus(gui.in_focus);
5736 
5737 	    if (im_activatekey_keyval != GDK_VoidSymbol)
5738 	    {
5739 		if (im_is_active)
5740 		{
5741 		    g_signal_handler_block(xic, im_commit_handler_id);
5742 		    im_synthesize_keypress(im_activatekey_keyval,
5743 						    im_activatekey_state);
5744 		    g_signal_handler_unblock(xic, im_commit_handler_id);
5745 		}
5746 	    }
5747 	    else
5748 	    {
5749 		im_shutdown();
5750 		xim_init();
5751 		xim_set_focus(gui.in_focus);
5752 	    }
5753 	}
5754     }
5755 
5756     if (p_imst == IM_ON_THE_SPOT)
5757 	preedit_start_col = MAXCOL;
5758     xim_has_preediting = FALSE;
5759 }
5760 
5761     int
5762 xim_queue_key_press_event(GdkEventKey *event, int down)
5763 {
5764     if (down)
5765     {
5766 	/*
5767 	 * Workaround GTK2 XIM 'feature' that always converts keypad keys to
5768 	 * chars., even when not part of an IM sequence (ref. feature of
5769 	 * gdk/gdkkeyuni.c).
5770 	 * Flag any keypad keys that might represent a single char.
5771 	 * If this (on its own - i.e., not part of an IM sequence) is
5772 	 * committed while we're processing one of these keys, we can ignore
5773 	 * that commit and go ahead & process it ourselves.  That way we can
5774 	 * still distinguish keypad keys for use in mappings.
5775 	 * Also add GDK_space to make <S-Space> work.
5776 	 */
5777 	switch (event->keyval)
5778 	{
5779 	    case GDK_KP_Add:      xim_expected_char = '+';  break;
5780 	    case GDK_KP_Subtract: xim_expected_char = '-';  break;
5781 	    case GDK_KP_Divide:   xim_expected_char = '/';  break;
5782 	    case GDK_KP_Multiply: xim_expected_char = '*';  break;
5783 	    case GDK_KP_Decimal:  xim_expected_char = '.';  break;
5784 	    case GDK_KP_Equal:    xim_expected_char = '=';  break;
5785 	    case GDK_KP_0:	  xim_expected_char = '0';  break;
5786 	    case GDK_KP_1:	  xim_expected_char = '1';  break;
5787 	    case GDK_KP_2:	  xim_expected_char = '2';  break;
5788 	    case GDK_KP_3:	  xim_expected_char = '3';  break;
5789 	    case GDK_KP_4:	  xim_expected_char = '4';  break;
5790 	    case GDK_KP_5:	  xim_expected_char = '5';  break;
5791 	    case GDK_KP_6:	  xim_expected_char = '6';  break;
5792 	    case GDK_KP_7:	  xim_expected_char = '7';  break;
5793 	    case GDK_KP_8:	  xim_expected_char = '8';  break;
5794 	    case GDK_KP_9:	  xim_expected_char = '9';  break;
5795 	    case GDK_space:	  xim_expected_char = ' ';  break;
5796 	    default:		  xim_expected_char = NUL;
5797 	}
5798 	xim_ignored_char = FALSE;
5799     }
5800 
5801     /*
5802      * When typing fFtT, XIM may be activated. Thus it must pass
5803      * gtk_im_context_filter_keypress() in Normal mode.
5804      * And while doing :sh too.
5805      */
5806     if (xic != NULL && !p_imdisable
5807 		    && (State & (INSERT | CMDLINE | NORMAL | EXTERNCMD)) != 0)
5808     {
5809 	/*
5810 	 * Filter 'imactivatekey' and map it to CTRL-^.  This way, Vim is
5811 	 * always aware of the current status of IM, and can even emulate
5812 	 * the activation key for modules that don't support one.
5813 	 */
5814 	if (event->keyval == im_activatekey_keyval
5815 	     && (event->state & im_activatekey_state) == im_activatekey_state)
5816 	{
5817 	    unsigned int state_mask;
5818 
5819 	    /* Require the state of the 3 most used modifiers to match exactly.
5820 	     * Otherwise e.g. <S-C-space> would be unusable for other purposes
5821 	     * if the IM activate key is <S-space>. */
5822 	    state_mask  = im_activatekey_state;
5823 	    state_mask |= ((int)GDK_SHIFT_MASK | (int)GDK_CONTROL_MASK
5824 							| (int)GDK_MOD1_MASK);
5825 
5826 	    if ((event->state & state_mask) != im_activatekey_state)
5827 		return FALSE;
5828 
5829 	    /* Don't send it a second time on GDK_KEY_RELEASE. */
5830 	    if (event->type != GDK_KEY_PRESS)
5831 		return TRUE;
5832 
5833 	    if (map_to_exists_mode((char_u *)"", LANGMAP, FALSE))
5834 	    {
5835 		im_set_active(FALSE);
5836 
5837 		/* ":lmap" mappings exists, toggle use of mappings. */
5838 		State ^= LANGMAP;
5839 		if (State & LANGMAP)
5840 		{
5841 		    curbuf->b_p_iminsert = B_IMODE_NONE;
5842 		    State &= ~LANGMAP;
5843 		}
5844 		else
5845 		{
5846 		    curbuf->b_p_iminsert = B_IMODE_LMAP;
5847 		    State |= LANGMAP;
5848 		}
5849 		return TRUE;
5850 	    }
5851 
5852 	    return gtk_im_context_filter_keypress(xic, event);
5853 	}
5854 
5855 	/* Don't filter events through the IM context if IM isn't active
5856 	 * right now.  Unlike with GTK+ 1.2 we cannot rely on the IM module
5857 	 * not doing anything before the activation key was sent. */
5858 	if (im_activatekey_keyval == GDK_VoidSymbol || im_is_active)
5859 	{
5860 	    int imresult = gtk_im_context_filter_keypress(xic, event);
5861 
5862 	    if (p_imst == IM_ON_THE_SPOT)
5863 	    {
5864 		/* Some XIM send following sequence:
5865 		 * 1. preedited string.
5866 		 * 2. committed string.
5867 		 * 3. line changed key.
5868 		 * 4. preedited string.
5869 		 * 5. remove preedited string.
5870 		 * if 3, Vim can't move back the above line for 5.
5871 		 * thus, this part should not parse the key. */
5872 		if (!imresult && preedit_start_col != MAXCOL
5873 					    && event->keyval == GDK_Return)
5874 		{
5875 		    im_synthesize_keypress(GDK_Return, 0U);
5876 		    return FALSE;
5877 		}
5878 	    }
5879 
5880 	    /* If XIM tried to commit a keypad key as a single char.,
5881 	     * ignore it so we can use the keypad key 'raw', for mappings. */
5882 	    if (xim_expected_char != NUL && xim_ignored_char)
5883 		/* We had a keypad key, and XIM tried to thieve it */
5884 		return FALSE;
5885 
5886 	    /* This is supposed to fix a problem with iBus, that space
5887 	     * characters don't work in input mode. */
5888 	    xim_expected_char = NUL;
5889 
5890 	    /* Normal processing */
5891 	    return imresult;
5892 	}
5893     }
5894 
5895     return FALSE;
5896 }
5897 
5898     int
5899 im_get_status(void)
5900 {
5901 #  ifdef FEAT_EVAL
5902     if (USE_IMSTATUSFUNC)
5903 	return call_imstatusfunc();
5904 #  endif
5905     return im_is_active;
5906 }
5907 
5908     int
5909 preedit_get_status(void)
5910 {
5911     return preedit_is_active;
5912 }
5913 
5914     int
5915 im_is_preediting(void)
5916 {
5917     return xim_has_preediting;
5918 }
5919 
5920 # else /* !FEAT_GUI_GTK */
5921 
5922 static int	xim_is_active = FALSE;  /* XIM should be active in the current
5923 					   mode */
5924 static int	xim_has_focus = FALSE;	/* XIM is really being used for Vim */
5925 #  ifdef FEAT_GUI_X11
5926 static XIMStyle	input_style;
5927 static int	status_area_enabled = TRUE;
5928 #  endif
5929 
5930 /*
5931  * Switch using XIM on/off.  This is used by the code that changes "State".
5932  * When 'imactivatefunc' is defined use that function instead.
5933  */
5934     void
5935 im_set_active(int active_arg)
5936 {
5937     int active = active_arg;
5938 
5939     /* If 'imdisable' is set, XIM is never active. */
5940     if (p_imdisable)
5941 	active = FALSE;
5942     else if (input_style & XIMPreeditPosition)
5943 	/* There is a problem in switching XIM off when preediting is used,
5944 	 * and it is not clear how this can be solved.  For now, keep XIM on
5945 	 * all the time, like it was done in Vim 5.8. */
5946 	active = TRUE;
5947 
5948 #  if defined(FEAT_EVAL)
5949     if (USE_IMACTIVATEFUNC)
5950     {
5951 	if (active != im_get_status())
5952 	{
5953 	    call_imactivatefunc(active);
5954 	    xim_has_focus = active;
5955 	}
5956 	return;
5957     }
5958 #  endif
5959 
5960     if (xic == NULL)
5961 	return;
5962 
5963     /* Remember the active state, it is needed when Vim gets keyboard focus. */
5964     xim_is_active = active;
5965     xim_set_preedit();
5966 }
5967 
5968 /*
5969  * Adjust using XIM for gaining or losing keyboard focus.  Also called when
5970  * "xim_is_active" changes.
5971  */
5972     void
5973 xim_set_focus(int focus)
5974 {
5975     if (xic == NULL)
5976 	return;
5977 
5978     /*
5979      * XIM only gets focus when the Vim window has keyboard focus and XIM has
5980      * been set active for the current mode.
5981      */
5982     if (focus && xim_is_active)
5983     {
5984 	if (!xim_has_focus)
5985 	{
5986 	    xim_has_focus = TRUE;
5987 	    XSetICFocus(xic);
5988 	}
5989     }
5990     else
5991     {
5992 	if (xim_has_focus)
5993 	{
5994 	    xim_has_focus = FALSE;
5995 	    XUnsetICFocus(xic);
5996 	}
5997     }
5998 }
5999 
6000     void
6001 im_set_position(int row UNUSED, int col UNUSED)
6002 {
6003     xim_set_preedit();
6004 }
6005 
6006 /*
6007  * Set the XIM to the current cursor position.
6008  */
6009     void
6010 xim_set_preedit(void)
6011 {
6012     XVaNestedList attr_list;
6013     XRectangle spot_area;
6014     XPoint over_spot;
6015     int line_space;
6016 
6017     if (xic == NULL)
6018 	return;
6019 
6020     xim_set_focus(TRUE);
6021 
6022     if (!xim_has_focus)
6023     {
6024 	/* hide XIM cursor */
6025 	over_spot.x = 0;
6026 	over_spot.y = -100; /* arbitrary invisible position */
6027 	attr_list = (XVaNestedList) XVaCreateNestedList(0,
6028 							XNSpotLocation,
6029 							&over_spot,
6030 							NULL);
6031 	XSetICValues(xic, XNPreeditAttributes, attr_list, NULL);
6032 	XFree(attr_list);
6033 	return;
6034     }
6035 
6036     if (input_style & XIMPreeditPosition)
6037     {
6038 	if (xim_fg_color == INVALCOLOR)
6039 	{
6040 	    xim_fg_color = gui.def_norm_pixel;
6041 	    xim_bg_color = gui.def_back_pixel;
6042 	}
6043 	over_spot.x = TEXT_X(gui.col);
6044 	over_spot.y = TEXT_Y(gui.row);
6045 	spot_area.x = 0;
6046 	spot_area.y = 0;
6047 	spot_area.height = gui.char_height * Rows;
6048 	spot_area.width  = gui.char_width * Columns;
6049 	line_space = gui.char_height;
6050 	attr_list = (XVaNestedList) XVaCreateNestedList(0,
6051 					XNSpotLocation, &over_spot,
6052 					XNForeground, (Pixel) xim_fg_color,
6053 					XNBackground, (Pixel) xim_bg_color,
6054 					XNArea, &spot_area,
6055 					XNLineSpace, line_space,
6056 					NULL);
6057 	if (XSetICValues(xic, XNPreeditAttributes, attr_list, NULL))
6058 	    EMSG(_("E284: Cannot set IC values"));
6059 	XFree(attr_list);
6060     }
6061 }
6062 
6063 #  if defined(FEAT_GUI_X11)
6064 static char e_xim[] = N_("E285: Failed to create input context");
6065 #  endif
6066 
6067 #  if defined(FEAT_GUI_X11) || defined(PROTO)
6068 #   if defined(XtSpecificationRelease) && XtSpecificationRelease >= 6 && !defined(SUN_SYSTEM)
6069 #    define USE_X11R6_XIM
6070 #   endif
6071 
6072 static int xim_real_init(Window x11_window, Display *x11_display);
6073 
6074 
6075 #  ifdef USE_X11R6_XIM
6076 static void xim_destroy_cb(XIM im, XPointer client_data, XPointer call_data);
6077 
6078     static void
6079 xim_instantiate_cb(
6080     Display	*display,
6081     XPointer	client_data UNUSED,
6082     XPointer	call_data UNUSED)
6083 {
6084     Window	x11_window;
6085     Display	*x11_display;
6086 
6087 #   ifdef XIM_DEBUG
6088     xim_log("xim_instantiate_cb()\n");
6089 #   endif
6090 
6091     gui_get_x11_windis(&x11_window, &x11_display);
6092     if (display != x11_display)
6093 	return;
6094 
6095     xim_real_init(x11_window, x11_display);
6096     gui_set_shellsize(FALSE, FALSE, RESIZE_BOTH);
6097     if (xic != NULL)
6098 	XUnregisterIMInstantiateCallback(x11_display, NULL, NULL, NULL,
6099 					 xim_instantiate_cb, NULL);
6100 }
6101 
6102     static void
6103 xim_destroy_cb(
6104     XIM		im UNUSED,
6105     XPointer	client_data UNUSED,
6106     XPointer	call_data UNUSED)
6107 {
6108     Window	x11_window;
6109     Display	*x11_display;
6110 
6111 #   ifdef XIM_DEBUG
6112     xim_log("xim_destroy_cb()\n");
6113    #endif
6114     gui_get_x11_windis(&x11_window, &x11_display);
6115 
6116     xic = NULL;
6117     status_area_enabled = FALSE;
6118 
6119     gui_set_shellsize(FALSE, FALSE, RESIZE_BOTH);
6120 
6121     XRegisterIMInstantiateCallback(x11_display, NULL, NULL, NULL,
6122 				   xim_instantiate_cb, NULL);
6123 }
6124 #  endif
6125 
6126     void
6127 xim_init(void)
6128 {
6129     Window	x11_window;
6130     Display	*x11_display;
6131 
6132 #  ifdef XIM_DEBUG
6133     xim_log("xim_init()\n");
6134 #  endif
6135 
6136     gui_get_x11_windis(&x11_window, &x11_display);
6137 
6138     xic = NULL;
6139 
6140     if (xim_real_init(x11_window, x11_display))
6141 	return;
6142 
6143     gui_set_shellsize(FALSE, FALSE, RESIZE_BOTH);
6144 
6145 #  ifdef USE_X11R6_XIM
6146     XRegisterIMInstantiateCallback(x11_display, NULL, NULL, NULL,
6147 				   xim_instantiate_cb, NULL);
6148 #  endif
6149 }
6150 
6151     static int
6152 xim_real_init(Window x11_window, Display *x11_display)
6153 {
6154     int		i;
6155     char	*p,
6156 		*s,
6157 		*ns,
6158 		*end,
6159 		tmp[1024];
6160 #  define IMLEN_MAX 40
6161     char	buf[IMLEN_MAX + 7];
6162     XIM		xim = NULL;
6163     XIMStyles	*xim_styles;
6164     XIMStyle	this_input_style = 0;
6165     Boolean	found;
6166     XPoint	over_spot;
6167     XVaNestedList preedit_list, status_list;
6168 
6169     input_style = 0;
6170     status_area_enabled = FALSE;
6171 
6172     if (xic != NULL)
6173 	return FALSE;
6174 
6175     if (gui.rsrc_input_method != NULL && *gui.rsrc_input_method != NUL)
6176     {
6177 	strcpy(tmp, gui.rsrc_input_method);
6178 	for (ns = s = tmp; ns != NULL && *s != NUL;)
6179 	{
6180 	    s = (char *)skipwhite((char_u *)s);
6181 	    if (*s == NUL)
6182 		break;
6183 	    if ((ns = end = strchr(s, ',')) == NULL)
6184 		end = s + strlen(s);
6185 	    while (isspace(((char_u *)end)[-1]))
6186 		end--;
6187 	    *end = NUL;
6188 
6189 	    if (strlen(s) <= IMLEN_MAX)
6190 	    {
6191 		strcpy(buf, "@im=");
6192 		strcat(buf, s);
6193 		if ((p = XSetLocaleModifiers(buf)) != NULL && *p != NUL
6194 			&& (xim = XOpenIM(x11_display, NULL, NULL, NULL))
6195 								      != NULL)
6196 		    break;
6197 	    }
6198 
6199 	    s = ns + 1;
6200 	}
6201     }
6202 
6203     if (xim == NULL && (p = XSetLocaleModifiers("")) != NULL && *p != NUL)
6204 	xim = XOpenIM(x11_display, NULL, NULL, NULL);
6205 
6206     /* This is supposed to be useful to obtain characters through
6207      * XmbLookupString() without really using a XIM. */
6208     if (xim == NULL && (p = XSetLocaleModifiers("@im=none")) != NULL
6209 								 && *p != NUL)
6210 	xim = XOpenIM(x11_display, NULL, NULL, NULL);
6211 
6212     if (xim == NULL)
6213     {
6214 	/* Only give this message when verbose is set, because too many people
6215 	 * got this message when they didn't want to use a XIM. */
6216 	if (p_verbose > 0)
6217 	{
6218 	    verbose_enter();
6219 	    EMSG(_("E286: Failed to open input method"));
6220 	    verbose_leave();
6221 	}
6222 	return FALSE;
6223     }
6224 
6225 #  ifdef USE_X11R6_XIM
6226     {
6227 	XIMCallback destroy_cb;
6228 
6229 	destroy_cb.callback = xim_destroy_cb;
6230 	destroy_cb.client_data = NULL;
6231 	if (XSetIMValues(xim, XNDestroyCallback, &destroy_cb, NULL))
6232 	    EMSG(_("E287: Warning: Could not set destroy callback to IM"));
6233     }
6234 #  endif
6235 
6236     if (XGetIMValues(xim, XNQueryInputStyle, &xim_styles, NULL) || !xim_styles)
6237     {
6238 	EMSG(_("E288: input method doesn't support any style"));
6239 	XCloseIM(xim);
6240 	return FALSE;
6241     }
6242 
6243     found = False;
6244     strcpy(tmp, gui.rsrc_preedit_type_name);
6245     for (s = tmp; s && !found; )
6246     {
6247 	while (*s && isspace((unsigned char)*s))
6248 	    s++;
6249 	if (!*s)
6250 	    break;
6251 	if ((ns = end = strchr(s, ',')) != 0)
6252 	    ns++;
6253 	else
6254 	    end = s + strlen(s);
6255 	while (isspace((unsigned char)*end))
6256 	    end--;
6257 	*end = '\0';
6258 
6259 	if (!strcmp(s, "OverTheSpot"))
6260 	    this_input_style = (XIMPreeditPosition | XIMStatusArea);
6261 	else if (!strcmp(s, "OffTheSpot"))
6262 	    this_input_style = (XIMPreeditArea | XIMStatusArea);
6263 	else if (!strcmp(s, "Root"))
6264 	    this_input_style = (XIMPreeditNothing | XIMStatusNothing);
6265 
6266 	for (i = 0; (unsigned short)i < xim_styles->count_styles; i++)
6267 	{
6268 	    if (this_input_style == xim_styles->supported_styles[i])
6269 	    {
6270 		found = True;
6271 		break;
6272 	    }
6273 	}
6274 	if (!found)
6275 	    for (i = 0; (unsigned short)i < xim_styles->count_styles; i++)
6276 	    {
6277 		if ((xim_styles->supported_styles[i] & this_input_style)
6278 			== (this_input_style & ~XIMStatusArea))
6279 		{
6280 		    this_input_style &= ~XIMStatusArea;
6281 		    found = True;
6282 		    break;
6283 		}
6284 	    }
6285 
6286 	s = ns;
6287     }
6288     XFree(xim_styles);
6289 
6290     if (!found)
6291     {
6292 	/* Only give this message when verbose is set, because too many people
6293 	 * got this message when they didn't want to use a XIM. */
6294 	if (p_verbose > 0)
6295 	{
6296 	    verbose_enter();
6297 	    EMSG(_("E289: input method doesn't support my preedit type"));
6298 	    verbose_leave();
6299 	}
6300 	XCloseIM(xim);
6301 	return FALSE;
6302     }
6303 
6304     over_spot.x = TEXT_X(gui.col);
6305     over_spot.y = TEXT_Y(gui.row);
6306     input_style = this_input_style;
6307 
6308     /* A crash was reported when trying to pass gui.norm_font as XNFontSet,
6309      * thus that has been removed.  Hopefully the default works... */
6310 #  ifdef FEAT_XFONTSET
6311     if (gui.fontset != NOFONTSET)
6312     {
6313 	preedit_list = XVaCreateNestedList(0,
6314 				XNSpotLocation, &over_spot,
6315 				XNForeground, (Pixel)gui.def_norm_pixel,
6316 				XNBackground, (Pixel)gui.def_back_pixel,
6317 				XNFontSet, (XFontSet)gui.fontset,
6318 				NULL);
6319 	status_list = XVaCreateNestedList(0,
6320 				XNForeground, (Pixel)gui.def_norm_pixel,
6321 				XNBackground, (Pixel)gui.def_back_pixel,
6322 				XNFontSet, (XFontSet)gui.fontset,
6323 				NULL);
6324     }
6325     else
6326 #  endif
6327     {
6328 	preedit_list = XVaCreateNestedList(0,
6329 				XNSpotLocation, &over_spot,
6330 				XNForeground, (Pixel)gui.def_norm_pixel,
6331 				XNBackground, (Pixel)gui.def_back_pixel,
6332 				NULL);
6333 	status_list = XVaCreateNestedList(0,
6334 				XNForeground, (Pixel)gui.def_norm_pixel,
6335 				XNBackground, (Pixel)gui.def_back_pixel,
6336 				NULL);
6337     }
6338 
6339     xic = XCreateIC(xim,
6340 		    XNInputStyle, input_style,
6341 		    XNClientWindow, x11_window,
6342 		    XNFocusWindow, gui.wid,
6343 		    XNPreeditAttributes, preedit_list,
6344 		    XNStatusAttributes, status_list,
6345 		    NULL);
6346     XFree(status_list);
6347     XFree(preedit_list);
6348     if (xic != NULL)
6349     {
6350 	if (input_style & XIMStatusArea)
6351 	{
6352 	    xim_set_status_area();
6353 	    status_area_enabled = TRUE;
6354 	}
6355 	else
6356 	    gui_set_shellsize(FALSE, FALSE, RESIZE_BOTH);
6357     }
6358     else
6359     {
6360 	if (!is_not_a_term())
6361 	    EMSG(_(e_xim));
6362 	XCloseIM(xim);
6363 	return FALSE;
6364     }
6365 
6366     return TRUE;
6367 }
6368 
6369 #  endif /* FEAT_GUI_X11 */
6370 
6371 /*
6372  * Get IM status.  When IM is on, return TRUE.  Else return FALSE.
6373  * FIXME: This doesn't work correctly: Having focus doesn't always mean XIM is
6374  * active, when not having focus XIM may still be active (e.g., when using a
6375  * tear-off menu item).
6376  */
6377     int
6378 im_get_status(void)
6379 {
6380 #  ifdef FEAT_EVAL
6381     if (USE_IMSTATUSFUNC)
6382 	return call_imstatusfunc();
6383 #  endif
6384     return xim_has_focus;
6385 }
6386 
6387 # endif /* !FEAT_GUI_GTK */
6388 
6389 # if !defined(FEAT_GUI_GTK) || defined(PROTO)
6390 /*
6391  * Set up the status area.
6392  *
6393  * This should use a separate Widget, but that seems not possible, because
6394  * preedit_area and status_area should be set to the same window as for the
6395  * text input.  Unfortunately this means the status area pollutes the text
6396  * window...
6397  */
6398     void
6399 xim_set_status_area(void)
6400 {
6401     XVaNestedList preedit_list = 0, status_list = 0, list = 0;
6402     XRectangle pre_area, status_area;
6403 
6404     if (xic == NULL)
6405 	return;
6406 
6407     if (input_style & XIMStatusArea)
6408     {
6409 	if (input_style & XIMPreeditArea)
6410 	{
6411 	    XRectangle *needed_rect;
6412 
6413 	    /* to get status_area width */
6414 	    status_list = XVaCreateNestedList(0, XNAreaNeeded,
6415 					      &needed_rect, NULL);
6416 	    XGetICValues(xic, XNStatusAttributes, status_list, NULL);
6417 	    XFree(status_list);
6418 
6419 	    status_area.width = needed_rect->width;
6420 	}
6421 	else
6422 	    status_area.width = gui.char_width * Columns;
6423 
6424 	status_area.x = 0;
6425 	status_area.y = gui.char_height * Rows + gui.border_offset;
6426 	if (gui.which_scrollbars[SBAR_BOTTOM])
6427 	    status_area.y += gui.scrollbar_height;
6428 #ifdef FEAT_MENU
6429 	if (gui.menu_is_active)
6430 	    status_area.y += gui.menu_height;
6431 #endif
6432 	status_area.height = gui.char_height;
6433 	status_list = XVaCreateNestedList(0, XNArea, &status_area, NULL);
6434     }
6435     else
6436     {
6437 	status_area.x = 0;
6438 	status_area.y = gui.char_height * Rows + gui.border_offset;
6439 	if (gui.which_scrollbars[SBAR_BOTTOM])
6440 	    status_area.y += gui.scrollbar_height;
6441 #ifdef FEAT_MENU
6442 	if (gui.menu_is_active)
6443 	    status_area.y += gui.menu_height;
6444 #endif
6445 	status_area.width = 0;
6446 	status_area.height = gui.char_height;
6447     }
6448 
6449     if (input_style & XIMPreeditArea)   /* off-the-spot */
6450     {
6451 	pre_area.x = status_area.x + status_area.width;
6452 	pre_area.y = gui.char_height * Rows + gui.border_offset;
6453 	pre_area.width = gui.char_width * Columns - pre_area.x;
6454 	if (gui.which_scrollbars[SBAR_BOTTOM])
6455 	    pre_area.y += gui.scrollbar_height;
6456 #ifdef FEAT_MENU
6457 	if (gui.menu_is_active)
6458 	    pre_area.y += gui.menu_height;
6459 #endif
6460 	pre_area.height = gui.char_height;
6461 	preedit_list = XVaCreateNestedList(0, XNArea, &pre_area, NULL);
6462     }
6463     else if (input_style & XIMPreeditPosition)   /* over-the-spot */
6464     {
6465 	pre_area.x = 0;
6466 	pre_area.y = 0;
6467 	pre_area.height = gui.char_height * Rows;
6468 	pre_area.width = gui.char_width * Columns;
6469 	preedit_list = XVaCreateNestedList(0, XNArea, &pre_area, NULL);
6470     }
6471 
6472     if (preedit_list && status_list)
6473 	list = XVaCreateNestedList(0, XNPreeditAttributes, preedit_list,
6474 				   XNStatusAttributes, status_list, NULL);
6475     else if (preedit_list)
6476 	list = XVaCreateNestedList(0, XNPreeditAttributes, preedit_list,
6477 				   NULL);
6478     else if (status_list)
6479 	list = XVaCreateNestedList(0, XNStatusAttributes, status_list,
6480 				   NULL);
6481     else
6482 	list = NULL;
6483 
6484     if (list)
6485     {
6486 	XSetICValues(xic, XNVaNestedList, list, NULL);
6487 	XFree(list);
6488     }
6489     if (status_list)
6490 	XFree(status_list);
6491     if (preedit_list)
6492 	XFree(preedit_list);
6493 }
6494 
6495     int
6496 xim_get_status_area_height(void)
6497 {
6498     if (status_area_enabled)
6499 	return gui.char_height;
6500     return 0;
6501 }
6502 # endif
6503 
6504 #else /* !defined(FEAT_XIM) */
6505 
6506 # ifdef IME_WITHOUT_XIM
6507 static int im_was_set_active = FALSE;
6508 
6509     int
6510 im_get_status(void)
6511 {
6512 #  if defined(FEAT_MBYTE) && defined(FEAT_EVAL)
6513     if (USE_IMSTATUSFUNC)
6514 	return call_imstatusfunc();
6515 #  endif
6516     return im_was_set_active;
6517 }
6518 
6519     void
6520 im_set_active(int active_arg)
6521 {
6522 #  if defined(FEAT_MBYTE) && defined(FEAT_EVAL)
6523     int	    active = !p_imdisable && active_arg;
6524 
6525     if (USE_IMACTIVATEFUNC && active != im_get_status())
6526     {
6527 	call_imactivatefunc(active);
6528 	im_was_set_active = active;
6529     }
6530 #  endif
6531 }
6532 
6533 #  ifdef FEAT_GUI
6534     void
6535 im_set_position(int row UNUSED, int col UNUSED)
6536 {
6537 }
6538 #  endif
6539 # endif
6540 
6541 #endif /* FEAT_XIM */
6542 
6543 #if defined(FEAT_MBYTE) || defined(PROTO)
6544 
6545 /*
6546  * Setup "vcp" for conversion from "from" to "to".
6547  * The names must have been made canonical with enc_canonize().
6548  * vcp->vc_type must have been initialized to CONV_NONE.
6549  * Note: cannot be used for conversion from/to ucs-2 and ucs-4 (will use utf-8
6550  * instead).
6551  * Afterwards invoke with "from" and "to" equal to NULL to cleanup.
6552  * Return FAIL when conversion is not supported, OK otherwise.
6553  */
6554     int
6555 convert_setup(vimconv_T *vcp, char_u *from, char_u *to)
6556 {
6557     return convert_setup_ext(vcp, from, TRUE, to, TRUE);
6558 }
6559 
6560 /*
6561  * As convert_setup(), but only when from_unicode_is_utf8 is TRUE will all
6562  * "from" unicode charsets be considered utf-8.  Same for "to".
6563  */
6564     int
6565 convert_setup_ext(
6566     vimconv_T	*vcp,
6567     char_u	*from,
6568     int		from_unicode_is_utf8,
6569     char_u	*to,
6570     int		to_unicode_is_utf8)
6571 {
6572     int		from_prop;
6573     int		to_prop;
6574     int		from_is_utf8;
6575     int		to_is_utf8;
6576 
6577     /* Reset to no conversion. */
6578 # ifdef USE_ICONV
6579     if (vcp->vc_type == CONV_ICONV && vcp->vc_fd != (iconv_t)-1)
6580 	iconv_close(vcp->vc_fd);
6581 # endif
6582     vcp->vc_type = CONV_NONE;
6583     vcp->vc_factor = 1;
6584     vcp->vc_fail = FALSE;
6585 
6586     /* No conversion when one of the names is empty or they are equal. */
6587     if (from == NULL || *from == NUL || to == NULL || *to == NUL
6588 						     || STRCMP(from, to) == 0)
6589 	return OK;
6590 
6591     from_prop = enc_canon_props(from);
6592     to_prop = enc_canon_props(to);
6593     if (from_unicode_is_utf8)
6594 	from_is_utf8 = from_prop & ENC_UNICODE;
6595     else
6596 	from_is_utf8 = from_prop == ENC_UNICODE;
6597     if (to_unicode_is_utf8)
6598 	to_is_utf8 = to_prop & ENC_UNICODE;
6599     else
6600 	to_is_utf8 = to_prop == ENC_UNICODE;
6601 
6602     if ((from_prop & ENC_LATIN1) && to_is_utf8)
6603     {
6604 	/* Internal latin1 -> utf-8 conversion. */
6605 	vcp->vc_type = CONV_TO_UTF8;
6606 	vcp->vc_factor = 2;	/* up to twice as long */
6607     }
6608     else if ((from_prop & ENC_LATIN9) && to_is_utf8)
6609     {
6610 	/* Internal latin9 -> utf-8 conversion. */
6611 	vcp->vc_type = CONV_9_TO_UTF8;
6612 	vcp->vc_factor = 3;	/* up to three as long (euro sign) */
6613     }
6614     else if (from_is_utf8 && (to_prop & ENC_LATIN1))
6615     {
6616 	/* Internal utf-8 -> latin1 conversion. */
6617 	vcp->vc_type = CONV_TO_LATIN1;
6618     }
6619     else if (from_is_utf8 && (to_prop & ENC_LATIN9))
6620     {
6621 	/* Internal utf-8 -> latin9 conversion. */
6622 	vcp->vc_type = CONV_TO_LATIN9;
6623     }
6624 #ifdef WIN3264
6625     /* Win32-specific codepage <-> codepage conversion without iconv. */
6626     else if ((from_is_utf8 || encname2codepage(from) > 0)
6627 	    && (to_is_utf8 || encname2codepage(to) > 0))
6628     {
6629 	vcp->vc_type = CONV_CODEPAGE;
6630 	vcp->vc_factor = 2;	/* up to twice as long */
6631 	vcp->vc_cpfrom = from_is_utf8 ? 0 : encname2codepage(from);
6632 	vcp->vc_cpto = to_is_utf8 ? 0 : encname2codepage(to);
6633     }
6634 #endif
6635 #ifdef MACOS_CONVERT
6636     else if ((from_prop & ENC_MACROMAN) && (to_prop & ENC_LATIN1))
6637     {
6638 	vcp->vc_type = CONV_MAC_LATIN1;
6639     }
6640     else if ((from_prop & ENC_MACROMAN) && to_is_utf8)
6641     {
6642 	vcp->vc_type = CONV_MAC_UTF8;
6643 	vcp->vc_factor = 2;	/* up to twice as long */
6644     }
6645     else if ((from_prop & ENC_LATIN1) && (to_prop & ENC_MACROMAN))
6646     {
6647 	vcp->vc_type = CONV_LATIN1_MAC;
6648     }
6649     else if (from_is_utf8 && (to_prop & ENC_MACROMAN))
6650     {
6651 	vcp->vc_type = CONV_UTF8_MAC;
6652     }
6653 #endif
6654 # ifdef USE_ICONV
6655     else
6656     {
6657 	/* Use iconv() for conversion. */
6658 	vcp->vc_fd = (iconv_t)my_iconv_open(
6659 		to_is_utf8 ? (char_u *)"utf-8" : to,
6660 		from_is_utf8 ? (char_u *)"utf-8" : from);
6661 	if (vcp->vc_fd != (iconv_t)-1)
6662 	{
6663 	    vcp->vc_type = CONV_ICONV;
6664 	    vcp->vc_factor = 4;	/* could be longer too... */
6665 	}
6666     }
6667 # endif
6668     if (vcp->vc_type == CONV_NONE)
6669 	return FAIL;
6670 
6671     return OK;
6672 }
6673 
6674 #if defined(FEAT_GUI) || defined(AMIGA) || defined(WIN3264) \
6675 	|| defined(PROTO)
6676 /*
6677  * Do conversion on typed input characters in-place.
6678  * The input and output are not NUL terminated!
6679  * Returns the length after conversion.
6680  */
6681     int
6682 convert_input(char_u *ptr, int len, int maxlen)
6683 {
6684     return convert_input_safe(ptr, len, maxlen, NULL, NULL);
6685 }
6686 #endif
6687 
6688 /*
6689  * Like convert_input(), but when there is an incomplete byte sequence at the
6690  * end return that as an allocated string in "restp" and set "*restlenp" to
6691  * the length.  If "restp" is NULL it is not used.
6692  */
6693     int
6694 convert_input_safe(
6695     char_u	*ptr,
6696     int		len,
6697     int		maxlen,
6698     char_u	**restp,
6699     int		*restlenp)
6700 {
6701     char_u	*d;
6702     int		dlen = len;
6703     int		unconvertlen = 0;
6704 
6705     d = string_convert_ext(&input_conv, ptr, &dlen,
6706 					restp == NULL ? NULL : &unconvertlen);
6707     if (d != NULL)
6708     {
6709 	if (dlen <= maxlen)
6710 	{
6711 	    if (unconvertlen > 0)
6712 	    {
6713 		/* Move the unconverted characters to allocated memory. */
6714 		*restp = alloc(unconvertlen);
6715 		if (*restp != NULL)
6716 		    mch_memmove(*restp, ptr + len - unconvertlen, unconvertlen);
6717 		*restlenp = unconvertlen;
6718 	    }
6719 	    mch_memmove(ptr, d, dlen);
6720 	}
6721 	else
6722 	    /* result is too long, keep the unconverted text (the caller must
6723 	     * have done something wrong!) */
6724 	    dlen = len;
6725 	vim_free(d);
6726     }
6727     return dlen;
6728 }
6729 
6730 /*
6731  * Convert text "ptr[*lenp]" according to "vcp".
6732  * Returns the result in allocated memory and sets "*lenp".
6733  * When "lenp" is NULL, use NUL terminated strings.
6734  * Illegal chars are often changed to "?", unless vcp->vc_fail is set.
6735  * When something goes wrong, NULL is returned and "*lenp" is unchanged.
6736  */
6737     char_u *
6738 string_convert(
6739     vimconv_T	*vcp,
6740     char_u	*ptr,
6741     int		*lenp)
6742 {
6743     return string_convert_ext(vcp, ptr, lenp, NULL);
6744 }
6745 
6746 /*
6747  * Like string_convert(), but when "unconvlenp" is not NULL and there are is
6748  * an incomplete sequence at the end it is not converted and "*unconvlenp" is
6749  * set to the number of remaining bytes.
6750  */
6751     char_u *
6752 string_convert_ext(
6753     vimconv_T	*vcp,
6754     char_u	*ptr,
6755     int		*lenp,
6756     int		*unconvlenp)
6757 {
6758     char_u	*retval = NULL;
6759     char_u	*d;
6760     int		len;
6761     int		i;
6762     int		l;
6763     int		c;
6764 
6765     if (lenp == NULL)
6766 	len = (int)STRLEN(ptr);
6767     else
6768 	len = *lenp;
6769     if (len == 0)
6770 	return vim_strsave((char_u *)"");
6771 
6772     switch (vcp->vc_type)
6773     {
6774 	case CONV_TO_UTF8:	/* latin1 to utf-8 conversion */
6775 	    retval = alloc(len * 2 + 1);
6776 	    if (retval == NULL)
6777 		break;
6778 	    d = retval;
6779 	    for (i = 0; i < len; ++i)
6780 	    {
6781 		c = ptr[i];
6782 		if (c < 0x80)
6783 		    *d++ = c;
6784 		else
6785 		{
6786 		    *d++ = 0xc0 + ((unsigned)c >> 6);
6787 		    *d++ = 0x80 + (c & 0x3f);
6788 		}
6789 	    }
6790 	    *d = NUL;
6791 	    if (lenp != NULL)
6792 		*lenp = (int)(d - retval);
6793 	    break;
6794 
6795 	case CONV_9_TO_UTF8:	/* latin9 to utf-8 conversion */
6796 	    retval = alloc(len * 3 + 1);
6797 	    if (retval == NULL)
6798 		break;
6799 	    d = retval;
6800 	    for (i = 0; i < len; ++i)
6801 	    {
6802 		c = ptr[i];
6803 		switch (c)
6804 		{
6805 		    case 0xa4: c = 0x20ac; break;   /* euro */
6806 		    case 0xa6: c = 0x0160; break;   /* S hat */
6807 		    case 0xa8: c = 0x0161; break;   /* S -hat */
6808 		    case 0xb4: c = 0x017d; break;   /* Z hat */
6809 		    case 0xb8: c = 0x017e; break;   /* Z -hat */
6810 		    case 0xbc: c = 0x0152; break;   /* OE */
6811 		    case 0xbd: c = 0x0153; break;   /* oe */
6812 		    case 0xbe: c = 0x0178; break;   /* Y */
6813 		}
6814 		d += utf_char2bytes(c, d);
6815 	    }
6816 	    *d = NUL;
6817 	    if (lenp != NULL)
6818 		*lenp = (int)(d - retval);
6819 	    break;
6820 
6821 	case CONV_TO_LATIN1:	/* utf-8 to latin1 conversion */
6822 	case CONV_TO_LATIN9:	/* utf-8 to latin9 conversion */
6823 	    retval = alloc(len + 1);
6824 	    if (retval == NULL)
6825 		break;
6826 	    d = retval;
6827 	    for (i = 0; i < len; ++i)
6828 	    {
6829 		l = utf_ptr2len_len(ptr + i, len - i);
6830 		if (l == 0)
6831 		    *d++ = NUL;
6832 		else if (l == 1)
6833 		{
6834 		    int l_w = utf8len_tab_zero[ptr[i]];
6835 
6836 		    if (l_w == 0)
6837 		    {
6838 			/* Illegal utf-8 byte cannot be converted */
6839 			vim_free(retval);
6840 			return NULL;
6841 		    }
6842 		    if (unconvlenp != NULL && l_w > len - i)
6843 		    {
6844 			/* Incomplete sequence at the end. */
6845 			*unconvlenp = len - i;
6846 			break;
6847 		    }
6848 		    *d++ = ptr[i];
6849 		}
6850 		else
6851 		{
6852 		    c = utf_ptr2char(ptr + i);
6853 		    if (vcp->vc_type == CONV_TO_LATIN9)
6854 			switch (c)
6855 			{
6856 			    case 0x20ac: c = 0xa4; break;   /* euro */
6857 			    case 0x0160: c = 0xa6; break;   /* S hat */
6858 			    case 0x0161: c = 0xa8; break;   /* S -hat */
6859 			    case 0x017d: c = 0xb4; break;   /* Z hat */
6860 			    case 0x017e: c = 0xb8; break;   /* Z -hat */
6861 			    case 0x0152: c = 0xbc; break;   /* OE */
6862 			    case 0x0153: c = 0xbd; break;   /* oe */
6863 			    case 0x0178: c = 0xbe; break;   /* Y */
6864 			    case 0xa4:
6865 			    case 0xa6:
6866 			    case 0xa8:
6867 			    case 0xb4:
6868 			    case 0xb8:
6869 			    case 0xbc:
6870 			    case 0xbd:
6871 			    case 0xbe: c = 0x100; break; /* not in latin9 */
6872 			}
6873 		    if (!utf_iscomposing(c))	/* skip composing chars */
6874 		    {
6875 			if (c < 0x100)
6876 			    *d++ = c;
6877 			else if (vcp->vc_fail)
6878 			{
6879 			    vim_free(retval);
6880 			    return NULL;
6881 			}
6882 			else
6883 			{
6884 			    *d++ = 0xbf;
6885 			    if (utf_char2cells(c) > 1)
6886 				*d++ = '?';
6887 			}
6888 		    }
6889 		    i += l - 1;
6890 		}
6891 	    }
6892 	    *d = NUL;
6893 	    if (lenp != NULL)
6894 		*lenp = (int)(d - retval);
6895 	    break;
6896 
6897 # ifdef MACOS_CONVERT
6898 	case CONV_MAC_LATIN1:
6899 	    retval = mac_string_convert(ptr, len, lenp, vcp->vc_fail,
6900 					'm', 'l', unconvlenp);
6901 	    break;
6902 
6903 	case CONV_LATIN1_MAC:
6904 	    retval = mac_string_convert(ptr, len, lenp, vcp->vc_fail,
6905 					'l', 'm', unconvlenp);
6906 	    break;
6907 
6908 	case CONV_MAC_UTF8:
6909 	    retval = mac_string_convert(ptr, len, lenp, vcp->vc_fail,
6910 					'm', 'u', unconvlenp);
6911 	    break;
6912 
6913 	case CONV_UTF8_MAC:
6914 	    retval = mac_string_convert(ptr, len, lenp, vcp->vc_fail,
6915 					'u', 'm', unconvlenp);
6916 	    break;
6917 # endif
6918 
6919 # ifdef USE_ICONV
6920 	case CONV_ICONV:	/* conversion with output_conv.vc_fd */
6921 	    retval = iconv_string(vcp, ptr, len, unconvlenp, lenp);
6922 	    break;
6923 # endif
6924 # ifdef WIN3264
6925 	case CONV_CODEPAGE:		/* codepage -> codepage */
6926 	{
6927 	    int		retlen;
6928 	    int		tmp_len;
6929 	    short_u	*tmp;
6930 
6931 	    /* 1. codepage/UTF-8  ->  ucs-2. */
6932 	    if (vcp->vc_cpfrom == 0)
6933 		tmp_len = utf8_to_utf16(ptr, len, NULL, NULL);
6934 	    else
6935 	    {
6936 		tmp_len = MultiByteToWideChar(vcp->vc_cpfrom,
6937 					unconvlenp ? MB_ERR_INVALID_CHARS : 0,
6938 					(char *)ptr, len, 0, 0);
6939 		if (tmp_len == 0
6940 			&& GetLastError() == ERROR_NO_UNICODE_TRANSLATION)
6941 		{
6942 		    if (lenp != NULL)
6943 			*lenp = 0;
6944 		    if (unconvlenp != NULL)
6945 			*unconvlenp = len;
6946 		    retval = alloc(1);
6947 		    if (retval)
6948 			retval[0] = NUL;
6949 		    return retval;
6950 		}
6951 	    }
6952 	    tmp = (short_u *)alloc(sizeof(short_u) * tmp_len);
6953 	    if (tmp == NULL)
6954 		break;
6955 	    if (vcp->vc_cpfrom == 0)
6956 		utf8_to_utf16(ptr, len, tmp, unconvlenp);
6957 	    else
6958 		MultiByteToWideChar(vcp->vc_cpfrom, 0,
6959 			(char *)ptr, len, tmp, tmp_len);
6960 
6961 	    /* 2. ucs-2  ->  codepage/UTF-8. */
6962 	    if (vcp->vc_cpto == 0)
6963 		retlen = utf16_to_utf8(tmp, tmp_len, NULL);
6964 	    else
6965 		retlen = WideCharToMultiByte(vcp->vc_cpto, 0,
6966 						    tmp, tmp_len, 0, 0, 0, 0);
6967 	    retval = alloc(retlen + 1);
6968 	    if (retval != NULL)
6969 	    {
6970 		if (vcp->vc_cpto == 0)
6971 		    utf16_to_utf8(tmp, tmp_len, retval);
6972 		else
6973 		    WideCharToMultiByte(vcp->vc_cpto, 0,
6974 					  tmp, tmp_len,
6975 					  (char *)retval, retlen, 0, 0);
6976 		retval[retlen] = NUL;
6977 		if (lenp != NULL)
6978 		    *lenp = retlen;
6979 	    }
6980 	    vim_free(tmp);
6981 	    break;
6982 	}
6983 # endif
6984     }
6985 
6986     return retval;
6987 }
6988 #endif
6989