Renovate display of non-ASCII messages on Windows.

nmisch · nmisch · commit 5f538ad004aa · 2013-06-26T11:17:33.000-04:00
GNU gettext selects a default encoding for the messages it emits in a
platform-specific manner; it uses the Windows ANSI code page on Windows
and follows LC_CTYPE on other platforms.  This is inconvenient for
PostgreSQL server processes, so realize consistent cross-platform
behavior by calling bind_textdomain_codeset() on Windows each time we
permanently change LC_CTYPE.  This primarily affects SQL_ASCII databases
and processes like the postmaster that do not attach to a database,
making their behavior consistent with PostgreSQL on non-Windows
platforms.  Messages from SQL_ASCII databases use the encoding implied
by the database LC_CTYPE, and messages from non-database processes use
LC_CTYPE from the postmaster system environment.  PlatformEncoding
becomes unused, so remove it.

Make write_console() prefer WriteConsoleW() to write() regardless of the
encodings in use.  In this situation, write() will invariably mishandle
non-ASCII characters.

elog.c has assumed that messages conform to the database encoding.
While usually true, this does not hold for SQL_ASCII and MULE_INTERNAL.
Introduce MessageEncoding to track the actual encoding of message text.
The present consumers are Windows-specific code for converting messages
to UTF16 for use in system interfaces.  This fixes the appearance in
Windows event logs and consoles of translated messages from SQL_ASCII
processes like the postmaster.  Note that SQL_ASCII inherently disclaims
a strong notion of encoding, so non-ASCII byte sequences interpolated
into messages by %s may yet yield a nonsensical message.  MULE_INTERNAL
has similar problems at present, albeit for a different reason: its lack
of libiconv support or a conversion to UTF8.

Consequently, one need no longer restart Windows with a different
Windows ANSI code page to broadly test backend logging under a given
language.  Changing the user's locale ("Format") is enough.  Several
accounts can simultaneously run postmasters under different locales, all
correctly logging localized messages to Windows event logs and consoles.

Alexander Law and Noah Misch
diff --git a/src/backend/main/main.c b/src/backend/main/main.c
@@ -265,6 +265,10 @@ startup_hacks(const char *progname)
 /*
  * Help display should match the options accepted by PostmasterMain()
  * and PostgresMain().
+ *
+ * XXX On Windows, non-ASCII localizations of these messages only display
+ * correctly if the console output code page covers the necessary characters.
+ * Messages emitted in write_console() do not exhibit this problem.
  */
 static void
 help(const char *progname)
diff --git a/src/backend/utils/adt/pg_locale.c b/src/backend/utils/adt/pg_locale.c
@@ -131,14 +131,16 @@ static char *IsoLocaleName(const char *);		/* MSVC specific */
 /*
  * pg_perm_setlocale
  *
- * This is identical to the libc function setlocale(), with the addition
- * that if the operation is successful, the corresponding LC_XXX environment
- * variable is set to match.  By setting the environment variable, we ensure
- * that any subsequent use of setlocale(..., "") will preserve the settings
- * made through this routine.  Of course, LC_ALL must also be unset to fully
- * ensure that, but that has to be done elsewhere after all the individual
- * LC_XXX variables have been set correctly.  (Thank you Perl for making this
- * kluge necessary.)
+ * This wraps the libc function setlocale(), with two additions.  First, when
+ * changing LC_CTYPE, update gettext's encoding for the current message
+ * domain.  GNU gettext automatically tracks LC_CTYPE on most platforms, but
+ * not on Windows.  Second, if the operation is successful, the corresponding
+ * LC_XXX environment variable is set to match.  By setting the environment
+ * variable, we ensure that any subsequent use of setlocale(..., "") will
+ * preserve the settings made through this routine.  Of course, LC_ALL must
+ * also be unset to fully ensure that, but that has to be done elsewhere after
+ * all the individual LC_XXX variables have been set correctly.  (Thank you
+ * Perl for making this kluge necessary.)
  */
 char *
 pg_perm_setlocale(int category, const char *locale)
@@ -172,6 +174,22 @@ pg_perm_setlocale(int category, const char *locale)
 	if (result == NULL)
 		return result;			/* fall out immediately on failure */
 
+	/*
+	 * Use the right encoding in translated messages.  Under ENABLE_NLS, let
+	 * pg_bind_textdomain_codeset() figure it out.  Under !ENABLE_NLS, message
+	 * format strings are ASCII, but database-encoding strings may enter the
+	 * message via %s.  This makes the overall message encoding equal to the
+	 * database encoding.
+	 */
+	if (category == LC_CTYPE)
+	{
+#ifdef ENABLE_NLS
+		SetMessageEncoding(pg_bind_textdomain_codeset(textdomain(NULL)));
+#else
+		SetMessageEncoding(GetDatabaseEncoding());
+#endif
+	}
+
 	switch (category)
 	{
 		case LC_COLLATE:
diff --git a/src/backend/utils/error/elog.c b/src/backend/utils/error/elog.c
@@ -1813,6 +1813,22 @@ write_syslog(int level, const char *line)
 #endif   /* HAVE_SYSLOG */
 
 #ifdef WIN32
+/*
+ * Get the PostgreSQL equivalent of the Windows ANSI code page.  "ANSI" system
+ * interfaces (e.g. CreateFileA()) expect string arguments in this encoding.
+ * Every process in a given system will find the same value at all times.
+ */
+static int
+GetACPEncoding(void)
+{
+	static int	encoding = -2;
+
+	if (encoding == -2)
+		encoding = pg_codepage_to_encoding(GetACP());
+
+	return encoding;
+}
+
 /*
  * Write a message line to the windows event log
  */
@@ -1858,16 +1874,18 @@ write_eventlog(int level, const char *line, int len)
 	}
 
 	/*
-	 * Convert message to UTF16 text and write it with ReportEventW, but
-	 * fall-back into ReportEventA if conversion failed.
+	 * If message character encoding matches the encoding expected by
+	 * ReportEventA(), call it to avoid the hazards of conversion.  Otherwise,
+	 * try to convert the message to UTF16 and write it with ReportEventW().
+	 * Fall back on ReportEventA() if conversion failed.
 	 *
 	 * Also verify that we are not on our way into error recursion trouble due
-	 * to error messages thrown deep inside pgwin32_toUTF16().
+	 * to error messages thrown deep inside pgwin32_message_to_UTF16().
 	 */
-	if (GetDatabaseEncoding() != GetPlatformEncoding() &&
-		!in_error_recursion_trouble())
+	if (!in_error_recursion_trouble() &&
+		GetMessageEncoding() != GetACPEncoding())
 	{
-		utf16 = pgwin32_toUTF16(line, len, NULL);
+		utf16 = pgwin32_message_to_UTF16(line, len, NULL);
 		if (utf16)
 		{
 			ReportEventW(evtHandle,
@@ -1879,6 +1897,7 @@ write_eventlog(int level, const char *line, int len)
 						 0,
 						 (LPCWSTR *) &utf16,
 						 NULL);
+			/* XXX Try ReportEventA() when ReportEventW() fails? */
 
 			pfree(utf16);
 			return;
@@ -1904,22 +1923,30 @@ write_console(const char *line, int len)
 #ifdef WIN32
 
 	/*
-	 * WriteConsoleW() will fail if stdout is redirected, so just fall through
+	 * Try to convert the message to UTF16 and write it with WriteConsoleW().
+	 * Fall back on write() if anything fails.
+	 *
+	 * In contrast to write_eventlog(), don't skip straight to write() based
+	 * on the applicable encodings.  Unlike WriteConsoleW(), write() depends
+	 * on the suitability of the console output code page.  Since we put
+	 * stderr into binary mode in SubPostmasterMain(), write() skips the
+	 * necessary translation anyway.
+	 *
+	 * WriteConsoleW() will fail if stderr is redirected, so just fall through
 	 * to writing unconverted to the logfile in this case.
 	 *
 	 * Since we palloc the structure required for conversion, also fall
 	 * through to writing unconverted if we have not yet set up
 	 * CurrentMemoryContext.
 	 */
-	if (GetDatabaseEncoding() != GetPlatformEncoding() &&
-		!in_error_recursion_trouble() &&
+	if (!in_error_recursion_trouble() &&
 		!redirection_done &&
 		CurrentMemoryContext != NULL)
 	{
 		WCHAR	   *utf16;
 		int			utf16len;
 
-		utf16 = pgwin32_toUTF16(line, len, &utf16len);
+		utf16 = pgwin32_message_to_UTF16(line, len, &utf16len);
 		if (utf16 != NULL)
 		{
 			HANDLE		stdHandle;
diff --git a/src/backend/utils/init/postinit.c b/src/backend/utils/init/postinit.c
@@ -357,11 +357,6 @@ CheckMyDatabase(const char *name, bool am_superuser)
 	SetConfigOption("lc_collate", collate, PGC_INTERNAL, PGC_S_OVERRIDE);
 	SetConfigOption("lc_ctype", ctype, PGC_INTERNAL, PGC_S_OVERRIDE);
 
-	/* Use the right encoding in translated messages */
-#ifdef ENABLE_NLS
-	pg_bind_textdomain_codeset(textdomain(NULL));
-#endif
-
 	ReleaseSysCache(tup);
 }
 
diff --git a/src/backend/utils/mb/encnames.c b/src/backend/utils/mb/encnames.c
@@ -352,10 +352,13 @@ pg_enc2name pg_enc2name_tbl[] =
 
 /* ----------
  * These are encoding names for gettext.
+ *
+ * This covers all encodings except MULE_INTERNAL, which is alien to gettext.
  * ----------
  */
 pg_enc2gettext pg_enc2gettext_tbl[] =
 {
+	{PG_SQL_ASCII, "US-ASCII"},
 	{PG_UTF8, "UTF-8"},
 	{PG_LATIN1, "LATIN1"},
 	{PG_LATIN2, "LATIN2"},
@@ -389,6 +392,13 @@ pg_enc2gettext pg_enc2gettext_tbl[] =
 	{PG_EUC_KR, "EUC-KR"},
 	{PG_EUC_TW, "EUC-TW"},
 	{PG_EUC_JIS_2004, "EUC-JP"},
+	{PG_SJIS, "SHIFT-JIS"},
+	{PG_BIG5, "BIG5"},
+	{PG_GBK, "GBK"},
+	{PG_UHC, "UHC"},
+	{PG_GB18030, "GB18030"},
+	{PG_JOHAB, "JOHAB"},
+	{PG_SHIFT_JIS_2004, "SHIFT_JISX0213"},
 	{0, NULL}
 };
 
diff --git a/src/backend/utils/mb/mbutils.c b/src/backend/utils/mb/mbutils.c
@@ -53,11 +53,11 @@ static FmgrInfo *ToServerConvProc = NULL;
 static FmgrInfo *ToClientConvProc = NULL;
 
 /*
- * These variables track the currently selected FE and BE encodings.
+ * These variables track the currently-selected encodings.
  */
 static pg_enc2name *ClientEncoding = &pg_enc2name_tbl[PG_SQL_ASCII];
 static pg_enc2name *DatabaseEncoding = &pg_enc2name_tbl[PG_SQL_ASCII];
-static pg_enc2name *PlatformEncoding = NULL;
+static pg_enc2name *MessageEncoding = &pg_enc2name_tbl[PG_SQL_ASCII];
 
 /*
  * During backend startup we can't set client encoding because we (a)
@@ -881,46 +881,102 @@ SetDatabaseEncoding(int encoding)
 	Assert(DatabaseEncoding->encoding == encoding);
 }
 
-/*
- * Bind gettext to the codeset equivalent with the database encoding.
- */
 void
-pg_bind_textdomain_codeset(const char *domainname)
+SetMessageEncoding(int encoding)
 {
-#if defined(ENABLE_NLS)
-	int			encoding = GetDatabaseEncoding();
-	int			i;
+	/* Some calls happen before we can elog()! */
+	Assert(PG_VALID_ENCODING(encoding));
 
-	/*
-	 * gettext() uses the codeset specified by LC_CTYPE by default, so if that
-	 * matches the database encoding we don't need to do anything. In CREATE
-	 * DATABASE, we enforce or trust that the locale's codeset matches
-	 * database encoding, except for the C locale. In C locale, we bind
-	 * gettext() explicitly to the right codeset.
-	 *
-	 * On Windows, though, gettext() tends to get confused so we always bind
-	 * it.
-	 */
-#ifndef WIN32
-	const char *ctype = setlocale(LC_CTYPE, NULL);
+	MessageEncoding = &pg_enc2name_tbl[encoding];
+	Assert(MessageEncoding->encoding == encoding);
+}
 
-	if (pg_strcasecmp(ctype, "C") != 0 && pg_strcasecmp(ctype, "POSIX") != 0)
-		return;
-#endif
+#ifdef ENABLE_NLS
+/*
+ * Make one bind_textdomain_codeset() call, translating a pg_enc to a gettext
+ * codeset.  Fails for MULE_INTERNAL, an encoding unknown to gettext; can also
+ * fail for gettext-internal causes like out-of-memory.
+ */
+static bool
+raw_pg_bind_textdomain_codeset(const char *domainname, int encoding)
+{
+	bool		elog_ok = (CurrentMemoryContext != NULL);
+	int			i;
 
 	for (i = 0; pg_enc2gettext_tbl[i].name != NULL; i++)
 	{
 		if (pg_enc2gettext_tbl[i].encoding == encoding)
 		{
 			if (bind_textdomain_codeset(domainname,
-										pg_enc2gettext_tbl[i].name) == NULL)
+										pg_enc2gettext_tbl[i].name) != NULL)
+				return true;
+
+			if (elog_ok)
 				elog(LOG, "bind_textdomain_codeset failed");
+			else
+				write_stderr("bind_textdomain_codeset failed");
+
 			break;
 		}
 	}
+
+	return false;
+}
+
+/*
+ * Bind a gettext message domain to the codeset corresponding to the database
+ * encoding.  For SQL_ASCII, instead bind to the codeset implied by LC_CTYPE.
+ * Return the MessageEncoding implied by the new settings.
+ *
+ * On most platforms, gettext defaults to the codeset implied by LC_CTYPE.
+ * When that matches the database encoding, we don't need to do anything.  In
+ * CREATE DATABASE, we enforce or trust that the locale's codeset matches the
+ * database encoding, except for the C locale.  (On Windows, we also permit a
+ * discrepancy under the UTF8 encoding.)  For the C locale, explicitly bind
+ * gettext to the right codeset.
+ *
+ * On Windows, gettext defaults to the Windows ANSI code page.  This is a
+ * convenient departure for software that passes the strings to Windows ANSI
+ * APIs, but we don't do that.  Compel gettext to use database encoding or,
+ * failing that, the LC_CTYPE encoding as it would on other platforms.
+ *
+ * This function is called before elog() and palloc() are usable.
+ */
+int
+pg_bind_textdomain_codeset(const char *domainname)
+{
+	bool		elog_ok = (CurrentMemoryContext != NULL);
+	int			encoding = GetDatabaseEncoding();
+	int			new_msgenc;
+
+#ifndef WIN32
+	const char *ctype = setlocale(LC_CTYPE, NULL);
+
+	if (pg_strcasecmp(ctype, "C") == 0 || pg_strcasecmp(ctype, "POSIX") == 0)
 #endif
+		if (encoding != PG_SQL_ASCII &&
+			raw_pg_bind_textdomain_codeset(domainname, encoding))
+			return encoding;
+
+	new_msgenc = pg_get_encoding_from_locale(NULL, elog_ok);
+	if (new_msgenc < 0)
+		new_msgenc = PG_SQL_ASCII;
+
+#ifdef WIN32
+	if (!raw_pg_bind_textdomain_codeset(domainname, new_msgenc))
+		/* On failure, the old message encoding remains valid. */
+		return GetMessageEncoding();
+#endif
+
+	return new_msgenc;
 }
+#endif
 
+/*
+ * The database encoding, also called the server encoding, represents the
+ * encoding of data stored in text-like data types.  Affected types include
+ * cstring, text, varchar, name, xml, and json.
+ */
 int
 GetDatabaseEncoding(void)
 {
@@ -949,19 +1005,17 @@ pg_client_encoding(PG_FUNCTION_ARGS)
 	return DirectFunctionCall1(namein, CStringGetDatum(ClientEncoding->name));
 }
 
+/*
+ * gettext() returns messages in this encoding.  This often matches the
+ * database encoding, but it differs for SQL_ASCII databases, for processes
+ * not attached to a database, and under a database encoding lacking iconv
+ * support (MULE_INTERNAL).
+ */
 int
-GetPlatformEncoding(void)
+GetMessageEncoding(void)
 {
-	if (PlatformEncoding == NULL)
-	{
-		/* try to determine encoding of server's environment locale */
-		int			encoding = pg_get_encoding_from_locale("", true);
-
-		if (encoding < 0)
-			encoding = PG_SQL_ASCII;
-		PlatformEncoding = &pg_enc2name_tbl[encoding];
-	}
-	return PlatformEncoding->encoding;
+	Assert(MessageEncoding);
+	return MessageEncoding->encoding;
 }
 
 #ifdef WIN32
@@ -971,13 +1025,13 @@ GetPlatformEncoding(void)
  * is also passed to utf16len if not null. Returns NULL iff failed.
  */
 WCHAR *
-pgwin32_toUTF16(const char *str, int len, int *utf16len)
+pgwin32_message_to_UTF16(const char *str, int len, int *utf16len)
 {
 	WCHAR	   *utf16;
 	int			dstlen;
 	UINT		codepage;
 
-	codepage = pg_enc2name_tbl[GetDatabaseEncoding()].codepage;
+	codepage = pg_enc2name_tbl[GetMessageEncoding()].codepage;
 
 	/*
 	 * Use MultiByteToWideChar directly if there is a corresponding codepage,
@@ -994,7 +1048,7 @@ pgwin32_toUTF16(const char *str, int len, int *utf16len)
 		char	   *utf8;
 
 		utf8 = (char *) pg_do_encoding_conversion((unsigned char *) str,
-										len, GetDatabaseEncoding(), PG_UTF8);
+										 len, GetMessageEncoding(), PG_UTF8);
 		if (utf8 != str)
 			len = strlen(utf8);
 
diff --git a/src/include/mb/pg_wchar.h b/src/include/mb/pg_wchar.h
diff --git a/src/include/port.h b/src/include/port.h
diff --git a/src/port/chklocale.c b/src/port/chklocale.c
