1 /*-------------------------------------------------------------------------
2  *
3  * chklocale.c
4  *		Functions for handling locale-related info
5  *
6  *
7  * Copyright (c) 1996-2017, PostgreSQL Global Development Group
8  *
9  *
10  * IDENTIFICATION
11  *	  src/port/chklocale.c
12  *
13  *-------------------------------------------------------------------------
14  */
15 
16 #ifndef FRONTEND
17 #include "postgres.h"
18 #else
19 #include "postgres_fe.h"
20 #endif
21 
22 #ifdef HAVE_LANGINFO_H
23 #include <langinfo.h>
24 #endif
25 
26 #include "mb/pg_wchar.h"
27 
28 
29 /*
30  * This table needs to recognize all the CODESET spellings for supported
31  * backend encodings, as well as frontend-only encodings where possible
32  * (the latter case is currently only needed for initdb to recognize
33  * error situations).  On Windows, we rely on entries for codepage
34  * numbers (CPnnn).
35  *
36  * Note that we search the table with pg_strcasecmp(), so variant
37  * capitalizations don't need their own entries.
38  */
39 struct encoding_match
40 {
41 	enum pg_enc pg_enc_code;
42 	const char *system_enc_name;
43 };
44 
45 static const struct encoding_match encoding_match_list[] = {
46 	{PG_EUC_JP, "EUC-JP"},
47 	{PG_EUC_JP, "eucJP"},
48 	{PG_EUC_JP, "IBM-eucJP"},
49 	{PG_EUC_JP, "sdeckanji"},
50 	{PG_EUC_JP, "CP20932"},
51 
52 	{PG_EUC_CN, "EUC-CN"},
53 	{PG_EUC_CN, "eucCN"},
54 	{PG_EUC_CN, "IBM-eucCN"},
55 	{PG_EUC_CN, "GB2312"},
56 	{PG_EUC_CN, "dechanzi"},
57 	{PG_EUC_CN, "CP20936"},
58 
59 	{PG_EUC_KR, "EUC-KR"},
60 	{PG_EUC_KR, "eucKR"},
61 	{PG_EUC_KR, "IBM-eucKR"},
62 	{PG_EUC_KR, "deckorean"},
63 	{PG_EUC_KR, "5601"},
64 	{PG_EUC_KR, "CP51949"},
65 
66 	{PG_EUC_TW, "EUC-TW"},
67 	{PG_EUC_TW, "eucTW"},
68 	{PG_EUC_TW, "IBM-eucTW"},
69 	{PG_EUC_TW, "cns11643"},
70 	/* No codepage for EUC-TW ? */
71 
72 	{PG_UTF8, "UTF-8"},
73 	{PG_UTF8, "utf8"},
74 	{PG_UTF8, "CP65001"},
75 
76 	{PG_LATIN1, "ISO-8859-1"},
77 	{PG_LATIN1, "ISO8859-1"},
78 	{PG_LATIN1, "iso88591"},
79 	{PG_LATIN1, "CP28591"},
80 
81 	{PG_LATIN2, "ISO-8859-2"},
82 	{PG_LATIN2, "ISO8859-2"},
83 	{PG_LATIN2, "iso88592"},
84 	{PG_LATIN2, "CP28592"},
85 
86 	{PG_LATIN3, "ISO-8859-3"},
87 	{PG_LATIN3, "ISO8859-3"},
88 	{PG_LATIN3, "iso88593"},
89 	{PG_LATIN3, "CP28593"},
90 
91 	{PG_LATIN4, "ISO-8859-4"},
92 	{PG_LATIN4, "ISO8859-4"},
93 	{PG_LATIN4, "iso88594"},
94 	{PG_LATIN4, "CP28594"},
95 
96 	{PG_LATIN5, "ISO-8859-9"},
97 	{PG_LATIN5, "ISO8859-9"},
98 	{PG_LATIN5, "iso88599"},
99 	{PG_LATIN5, "CP28599"},
100 
101 	{PG_LATIN6, "ISO-8859-10"},
102 	{PG_LATIN6, "ISO8859-10"},
103 	{PG_LATIN6, "iso885910"},
104 
105 	{PG_LATIN7, "ISO-8859-13"},
106 	{PG_LATIN7, "ISO8859-13"},
107 	{PG_LATIN7, "iso885913"},
108 
109 	{PG_LATIN8, "ISO-8859-14"},
110 	{PG_LATIN8, "ISO8859-14"},
111 	{PG_LATIN8, "iso885914"},
112 
113 	{PG_LATIN9, "ISO-8859-15"},
114 	{PG_LATIN9, "ISO8859-15"},
115 	{PG_LATIN9, "iso885915"},
116 	{PG_LATIN9, "CP28605"},
117 
118 	{PG_LATIN10, "ISO-8859-16"},
119 	{PG_LATIN10, "ISO8859-16"},
120 	{PG_LATIN10, "iso885916"},
121 
122 	{PG_KOI8R, "KOI8-R"},
123 	{PG_KOI8R, "CP20866"},
124 
125 	{PG_KOI8U, "KOI8-U"},
126 	{PG_KOI8U, "CP21866"},
127 
128 	{PG_WIN866, "CP866"},
129 	{PG_WIN874, "CP874"},
130 	{PG_WIN1250, "CP1250"},
131 	{PG_WIN1251, "CP1251"},
132 	{PG_WIN1251, "ansi-1251"},
133 	{PG_WIN1252, "CP1252"},
134 	{PG_WIN1253, "CP1253"},
135 	{PG_WIN1254, "CP1254"},
136 	{PG_WIN1255, "CP1255"},
137 	{PG_WIN1256, "CP1256"},
138 	{PG_WIN1257, "CP1257"},
139 	{PG_WIN1258, "CP1258"},
140 
141 	{PG_ISO_8859_5, "ISO-8859-5"},
142 	{PG_ISO_8859_5, "ISO8859-5"},
143 	{PG_ISO_8859_5, "iso88595"},
144 	{PG_ISO_8859_5, "CP28595"},
145 
146 	{PG_ISO_8859_6, "ISO-8859-6"},
147 	{PG_ISO_8859_6, "ISO8859-6"},
148 	{PG_ISO_8859_6, "iso88596"},
149 	{PG_ISO_8859_6, "CP28596"},
150 
151 	{PG_ISO_8859_7, "ISO-8859-7"},
152 	{PG_ISO_8859_7, "ISO8859-7"},
153 	{PG_ISO_8859_7, "iso88597"},
154 	{PG_ISO_8859_7, "CP28597"},
155 
156 	{PG_ISO_8859_8, "ISO-8859-8"},
157 	{PG_ISO_8859_8, "ISO8859-8"},
158 	{PG_ISO_8859_8, "iso88598"},
159 	{PG_ISO_8859_8, "CP28598"},
160 
161 	{PG_SJIS, "SJIS"},
162 	{PG_SJIS, "PCK"},
163 	{PG_SJIS, "CP932"},
164 	{PG_SJIS, "SHIFT_JIS"},
165 
166 	{PG_BIG5, "BIG5"},
167 	{PG_BIG5, "BIG5HKSCS"},
168 	{PG_BIG5, "Big5-HKSCS"},
169 	{PG_BIG5, "CP950"},
170 
171 	{PG_GBK, "GBK"},
172 	{PG_GBK, "CP936"},
173 
174 	{PG_UHC, "UHC"},
175 	{PG_UHC, "CP949"},
176 
177 	{PG_JOHAB, "JOHAB"},
178 	{PG_JOHAB, "CP1361"},
179 
180 	{PG_GB18030, "GB18030"},
181 	{PG_GB18030, "CP54936"},
182 
183 	{PG_SHIFT_JIS_2004, "SJIS_2004"},
184 
185 	{PG_SQL_ASCII, "US-ASCII"},
186 
187 	{PG_SQL_ASCII, NULL}		/* end marker */
188 };
189 
190 #ifdef WIN32
191 /*
192  * On Windows, use CP<code page number> instead of the nl_langinfo() result
193  *
194  * Visual Studio 2012 expanded the set of valid LC_CTYPE values, so have its
195  * locale machinery determine the code page.  See comments at IsoLocaleName().
196  * For other compilers, follow the locale's predictable format.
197  *
198  * Visual Studio 2015 should still be able to do the same, but the declaration
199  * of lc_codepage is missing in _locale_t, causing this code compilation to
200  * fail, hence this falls back instead on GetLocaleInfoEx. VS 2015 may be an
201  * exception and post-VS2015 versions should be able to handle properly the
202  * codepage number using _create_locale(). So, instead of the same logic as
203  * VS 2012 and VS 2013, this routine uses GetLocaleInfoEx to parse short
204  * locale names like "de-DE", "fr-FR", etc. If those cannot be parsed correctly
205  * process falls back to the pre-VS-2010 manual parsing done with
206  * using <Language>_<Country>.<CodePage> as a base.
207  *
208  * Returns a malloc()'d string for the caller to free.
209  */
210 static char *
win32_langinfo(const char * ctype)211 win32_langinfo(const char *ctype)
212 {
213 	char	   *r = NULL;
214 
215 #if (_MSC_VER >= 1700) && (_MSC_VER < 1900)
216 	_locale_t	loct = NULL;
217 
218 	loct = _create_locale(LC_CTYPE, ctype);
219 	if (loct != NULL)
220 	{
221 		r = malloc(16);			/* excess */
222 		if (r != NULL)
223 			sprintf(r, "CP%u", loct->locinfo->lc_codepage);
224 		_free_locale(loct);
225 	}
226 #else
227 	char	   *codepage;
228 
229 #if (_MSC_VER >= 1900)
230 	uint32		cp;
231 	WCHAR		wctype[LOCALE_NAME_MAX_LENGTH];
232 
233 	memset(wctype, 0, sizeof(wctype));
234 	MultiByteToWideChar(CP_ACP, 0, ctype, -1, wctype, LOCALE_NAME_MAX_LENGTH);
235 
236 	if (GetLocaleInfoEx(wctype,
237 						LOCALE_IDEFAULTANSICODEPAGE | LOCALE_RETURN_NUMBER,
238 						(LPWSTR) &cp, sizeof(cp) / sizeof(WCHAR)) > 0)
239 	{
240 		r = malloc(16);			/* excess */
241 		if (r != NULL)
242 		{
243 			/*
244 			 * If the return value is CP_ACP that means no ANSI code page is
245 			 * available, so only Unicode can be used for the locale.
246 			 */
247 			if (cp == CP_ACP)
248 				strcpy(r, "utf8");
249 			else
250 				sprintf(r, "CP%u", cp);
251 		}
252 	}
253 	else
254 #endif
255 	{
256 		/*
257 		 * Locale format on Win32 is <Language>_<Country>.<CodePage>.  For
258 		 * example, English_United States.1252.  If we see digits after the
259 		 * last dot, assume it's a codepage number.  Otherwise, we might be
260 		 * dealing with a Unix-style locale string; Windows' setlocale() will
261 		 * take those even though GetLocaleInfoEx() won't, so we end up here.
262 		 * In that case, just return what's after the last dot and hope we can
263 		 * find it in our table.
264 		 */
265 		codepage = strrchr(ctype, '.');
266 		if (codepage != NULL)
267 		{
268 			size_t		ln;
269 
270 			codepage++;
271 			ln = strlen(codepage);
272 			r = malloc(ln + 3);
273 			if (r != NULL)
274 			{
275 				if (strspn(codepage, "0123456789") == ln)
276 					sprintf(r, "CP%s", codepage);
277 				else
278 					strcpy(r, codepage);
279 			}
280 		}
281 
282 	}
283 #endif
284 
285 	return r;
286 }
287 
288 #ifndef FRONTEND
289 /*
290  * Given a Windows code page identifier, find the corresponding PostgreSQL
291  * encoding.  Issue a warning and return -1 if none found.
292  */
293 int
pg_codepage_to_encoding(UINT cp)294 pg_codepage_to_encoding(UINT cp)
295 {
296 	char		sys[16];
297 	int			i;
298 
299 	sprintf(sys, "CP%u", cp);
300 
301 	/* Check the table */
302 	for (i = 0; encoding_match_list[i].system_enc_name; i++)
303 		if (pg_strcasecmp(sys, encoding_match_list[i].system_enc_name) == 0)
304 			return encoding_match_list[i].pg_enc_code;
305 
306 	ereport(WARNING,
307 			(errmsg("could not determine encoding for codeset \"%s\"", sys)));
308 
309 	return -1;
310 }
311 #endif
312 #endif							/* WIN32 */
313 
314 #if (defined(HAVE_LANGINFO_H) && defined(CODESET)) || defined(WIN32)
315 
316 /*
317  * Given a setting for LC_CTYPE, return the Postgres ID of the associated
318  * encoding, if we can determine it.  Return -1 if we can't determine it.
319  *
320  * Pass in NULL to get the encoding for the current locale setting.
321  * Pass "" to get the encoding selected by the server's environment.
322  *
323  * If the result is PG_SQL_ASCII, callers should treat it as being compatible
324  * with any desired encoding.
325  *
326  * If running in the backend and write_message is false, this function must
327  * cope with the possibility that elog() and palloc() are not yet usable.
328  */
329 int
pg_get_encoding_from_locale(const char * ctype,bool write_message)330 pg_get_encoding_from_locale(const char *ctype, bool write_message)
331 {
332 	char	   *sys;
333 	int			i;
334 
335 	/* Get the CODESET property, and also LC_CTYPE if not passed in */
336 	if (ctype)
337 	{
338 		char	   *save;
339 		char	   *name;
340 
341 		/* If locale is C or POSIX, we can allow all encodings */
342 		if (pg_strcasecmp(ctype, "C") == 0 ||
343 			pg_strcasecmp(ctype, "POSIX") == 0)
344 			return PG_SQL_ASCII;
345 
346 		save = setlocale(LC_CTYPE, NULL);
347 		if (!save)
348 			return -1;			/* setlocale() broken? */
349 		/* must copy result, or it might change after setlocale */
350 		save = strdup(save);
351 		if (!save)
352 			return -1;			/* out of memory; unlikely */
353 
354 		name = setlocale(LC_CTYPE, ctype);
355 		if (!name)
356 		{
357 			free(save);
358 			return -1;			/* bogus ctype passed in? */
359 		}
360 
361 #ifndef WIN32
362 		sys = nl_langinfo(CODESET);
363 		if (sys)
364 			sys = strdup(sys);
365 #else
366 		sys = win32_langinfo(name);
367 #endif
368 
369 		setlocale(LC_CTYPE, save);
370 		free(save);
371 	}
372 	else
373 	{
374 		/* much easier... */
375 		ctype = setlocale(LC_CTYPE, NULL);
376 		if (!ctype)
377 			return -1;			/* setlocale() broken? */
378 
379 		/* If locale is C or POSIX, we can allow all encodings */
380 		if (pg_strcasecmp(ctype, "C") == 0 ||
381 			pg_strcasecmp(ctype, "POSIX") == 0)
382 			return PG_SQL_ASCII;
383 
384 #ifndef WIN32
385 		sys = nl_langinfo(CODESET);
386 		if (sys)
387 			sys = strdup(sys);
388 #else
389 		sys = win32_langinfo(ctype);
390 #endif
391 	}
392 
393 	if (!sys)
394 		return -1;				/* out of memory; unlikely */
395 
396 	/* Check the table */
397 	for (i = 0; encoding_match_list[i].system_enc_name; i++)
398 	{
399 		if (pg_strcasecmp(sys, encoding_match_list[i].system_enc_name) == 0)
400 		{
401 			free(sys);
402 			return encoding_match_list[i].pg_enc_code;
403 		}
404 	}
405 
406 	/* Special-case kluges for particular platforms go here */
407 
408 #ifdef __darwin__
409 
410 	/*
411 	 * Current macOS has many locales that report an empty string for CODESET,
412 	 * but they all seem to actually use UTF-8.
413 	 */
414 	if (strlen(sys) == 0)
415 	{
416 		free(sys);
417 		return PG_UTF8;
418 	}
419 #endif
420 
421 	/*
422 	 * We print a warning if we got a CODESET string but couldn't recognize
423 	 * it.  This means we need another entry in the table.
424 	 */
425 	if (write_message)
426 	{
427 #ifdef FRONTEND
428 		fprintf(stderr, _("could not determine encoding for locale \"%s\": codeset is \"%s\""),
429 				ctype, sys);
430 		/* keep newline separate so there's only one translatable string */
431 		fputc('\n', stderr);
432 #else
433 		ereport(WARNING,
434 				(errmsg("could not determine encoding for locale \"%s\": codeset is \"%s\"",
435 						ctype, sys)));
436 #endif
437 	}
438 
439 	free(sys);
440 	return -1;
441 }
442 #else							/* (HAVE_LANGINFO_H && CODESET) || WIN32 */
443 
444 /*
445  * stub if no multi-language platform support
446  *
447  * Note: we could return -1 here, but that would have the effect of
448  * forcing users to specify an encoding to initdb on such platforms.
449  * It seems better to silently default to SQL_ASCII.
450  */
451 int
pg_get_encoding_from_locale(const char * ctype,bool write_message)452 pg_get_encoding_from_locale(const char *ctype, bool write_message)
453 {
454 	return PG_SQL_ASCII;
455 }
456 
457 #endif							/* (HAVE_LANGINFO_H && CODESET) || WIN32 */
458