/////////////////////////////////////////////////////////////////////////////
// Name: strconv.cpp
// Purpose: Unicode conversion classes
-// Author: Ove Kaaven, Robert Roebling, Vadim Zeitlin
+// Author: Ove Kaaven, Robert Roebling, Vadim Zeitlin, Vaclav Slavik
// Modified by:
// Created: 29/01/98
// RCS-ID: $Id$
-// Copyright: (c) 1999 Ove Kaaven, Robert Roebling, Vadim Zeitlin
+// Copyright: (c) 1999 Ove Kaaven, Robert Roebling, Vadim Zeitlin, Vaclav Slavik
// Licence: wxWindows license
/////////////////////////////////////////////////////////////////////////////
#include <string.h>
#include <stdlib.h>
+
+#include "wx/debug.h"
+#include "wx/strconv.h"
+#include "wx/intl.h"
+#include "wx/log.h"
+
+// ----------------------------------------------------------------------------
+// globals
+// ----------------------------------------------------------------------------
+
+WXDLLEXPORT_DATA(wxMBConv *) wxConvCurrent = &wxConvLibc;
+
+
+// ============================================================================
+// implementation
+// ============================================================================
+
+#if wxUSE_WCHAR_T
+
#ifdef __SALFORDC__
#include <clib.h>
#endif
#include <windows.h>
#endif
-#include "wx/debug.h"
-#include "wx/strconv.h"
-#include "wx/intl.h"
-#include "wx/log.h"
-
-#if defined(WORDS_BIGENDIAN) || defined(__STDC_ISO_10646__)
-#define BSWAP_UCS4(str, len)
-#define BSWAP_UCS2(str, len)
-#else
#define BSWAP_UCS4(str, len) { unsigned _c; for (_c=0; _c<len; _c++) str[_c]=wxUINT32_SWAP_ALWAYS(str[_c]); }
-#define BSWAP_UCS2(str, len) { unsigned _c; for (_c=0; _c<len; _c++) str[_c]=wxUINT16_SWAP_ALWAYS(str[_c]); }
-#define WC_NEED_BSWAP
-#endif
-#define BSWAP_UTF32(str, len) BSWAP_UCS4(str, len)
-#define BSWAP_UTF16(str, len) BSWAP_UCS2(str, len)
+#define BSWAP_UTF16(str, len) { unsigned _c; for (_c=0; _c<len; _c++) str[_c]=wxUINT16_SWAP_ALWAYS(str[_c]); }
// under Unix SIZEOF_WCHAR_T is defined by configure, but under other platforms
// it might be not defined - assume the most common value
#endif // !defined(SIZEOF_WCHAR_T)
#if SIZEOF_WCHAR_T == 4
- #define WC_NAME "UCS4"
- #define WC_BSWAP BSWAP_UCS4
+ #define WC_NAME "UCS4"
+ #define WC_BSWAP BSWAP_UCS4
+ #ifdef WORDS_BIGENDIAN
+ #define WC_NAME_BEST "UCS-4BE"
+ #else
+ #define WC_NAME_BEST "UCS-4LE"
+ #endif
#elif SIZEOF_WCHAR_T == 2
- #define WC_NAME "UTF16"
- #define WC_BSWAP BSWAP_UTF16
+ #define WC_NAME "UTF16"
+ #define WC_BSWAP BSWAP_UTF16
#define WC_UTF16
+ #ifdef WORDS_BIGENDIAN
+ #define WC_NAME_BEST "UTF-16BE"
+ #else
+ #define WC_NAME_BEST "UTF-16LE"
+ #endif
#else // sizeof(wchar_t) != 2 nor 4
// I don't know what to do about this
#error "Weird sizeof(wchar_t): please report your platform details to wx-users mailing list"
#endif
-// ----------------------------------------------------------------------------
-// globals
-// ----------------------------------------------------------------------------
-
-WXDLLEXPORT_DATA(wxMBConv *) wxConvCurrent = &wxConvLibc;
-
-// ============================================================================
-// implementation
-// ============================================================================
-
-#if wxUSE_WCHAR_T
#ifdef WC_UTF16
#ifdef HAVE_ICONV_H
+bool g_wcNeedsSwap = FALSE;
+static const char *g_wcCharset = NULL;
+
// VS: glibc 2.1.3 is broken in that iconv() conversion to/from UCS4 fails with E2BIG
// if output buffer is _exactly_ as big as needed. Such case is (unless there's
// yet another bug in glibc) the only case when iconv() returns with (size_t)-1
IC_CharSet(const wxChar *name)
: wxCharacterSet(name)
{
- m2w = iconv_open(WC_NAME, wxConvLibc.cWX2MB(cname));
- w2m = iconv_open(wxConvLibc.cWX2MB(cname), WC_NAME);
- }
+ // check for charset that represents wchar_t:
+ if (g_wcCharset == NULL)
+ {
+ g_wcNeedsSwap = FALSE;
+
+ // try charset with explicit bytesex info (e.g. "UCS-4LE"):
+ g_wcCharset = WC_NAME_BEST;
+ m2w = iconv_open(g_wcCharset, wxConvLibc.cWX2MB(name));
- ~IC_CharSet()
+ if (m2w == (iconv_t)-1)
+ {
+ // try charset w/o bytesex info (e.g. "UCS4")
+ // and check for bytesex ourselves:
+ g_wcCharset = WC_NAME;
+ m2w = iconv_open(g_wcCharset, wxConvLibc.cWX2MB(name));
+
+ // last bet, try if it knows WCHAR_T pseudo-charset
+ if (m2w == (iconv_t)-1)
+ {
+ g_wcCharset = "WCHAR_T";
+ m2w = iconv_open(g_wcCharset, wxConvLibc.cWX2MB(name));
+ }
+
+ if (m2w != (iconv_t)-1)
+ {
+ char buf[2], *bufPtr;
+ wchar_t wbuf[2], *wbufPtr;
+ size_t insz, outsz;
+ size_t res;
+
+ buf[0] = 'A';
+ buf[1] = 0;
+ wbuf[0] = 0;
+ insz = 2;
+ outsz = SIZEOF_WCHAR_T * 2;
+ wbufPtr = wbuf;
+ bufPtr = buf;
+
+ #ifdef WX_ICONV_TAKES_CHAR
+ res = iconv(m2w, (char**)&bufPtr, &insz, (char**)&wbufPtr, &outsz);
+ #else
+ res = iconv(m2w, (const char**)&bufPtr, &insz, (char**)&wbufPtr, &outsz);
+ #endif
+ if (ICONV_FAILED(res, insz))
+ {
+ g_wcCharset = NULL;
+ wxLogLastError(wxT("iconv"));
+ wxLogError(_("Convertion to charset '%s' doesn't work."), name);
+ }
+ else
+ {
+ g_wcNeedsSwap = (wbuf[0] != (wchar_t)buf[0]);
+ }
+ }
+ else
+ {
+ g_wcCharset = NULL;
+ wxLogError(_("Don't know how to convert to/from charset '%s'."), name);
+ }
+ }
+ wxLogTrace(wxT("strconv"), wxT("wchar_t charset is '%s', needs swap: %i"), g_wcCharset, g_wcNeedsSwap);
+ }
+ else
+ m2w = iconv_open(g_wcCharset, wxConvLibc.cWX2MB(name));
+
+ w2m = iconv_open(wxConvLibc.cWX2MB(name), g_wcCharset);
+ }
+
+ ~IC_CharSet()
{
if ( m2w != (iconv_t)-1 )
iconv_close(m2w);
cres = iconv(m2w, &pszPtr, &inbuf, (char**)&bufPtr, &outbuf);
#endif
res = n - (outbuf / SIZEOF_WCHAR_T);
- // convert to native endianness
-#ifdef WC_NEED_BSWAP
- WC_BSWAP(buf /* _not_ bufPtr */, res)
-#endif
+
+ if (g_wcNeedsSwap)
+ {
+ // convert to native endianness
+ WC_BSWAP(buf /* _not_ bufPtr */, res)
+ }
}
else
{
}
if (ICONV_FAILED(cres, inbuf))
+ {
+ //VS: it is ok if iconv fails, hence trace only
+ wxLogTrace(wxT("strconv"), wxT("iconv failed: %s"), wxSysErrorMsg(wxSysErrorCode()));
return (size_t)-1;
+ }
return res;
}
#endif
size_t outbuf = n;
size_t res, cres;
+
+ wchar_t *tmpbuf = 0;
+
+ if (g_wcNeedsSwap)
+ {
+ // need to copy to temp buffer to switch endianness
+ // this absolutely doesn't rock!
+ // (no, doing WC_BSWAP twice on the original buffer won't help, as it
+ // could be in read-only memory, or be accessed in some other thread)
+ tmpbuf=(wchar_t*)malloc((inbuf+1)*SIZEOF_WCHAR_T);
+ memcpy(tmpbuf,psz,(inbuf+1)*SIZEOF_WCHAR_T);
+ WC_BSWAP(tmpbuf, inbuf)
+ psz=tmpbuf;
+ }
-#ifdef WC_NEED_BSWAP
- // need to copy to temp buffer to switch endianness
- // this absolutely doesn't rock!
- // (no, doing WC_BSWAP twice on the original buffer won't help, as it
- // could be in read-only memory, or be accessed in some other thread)
- wchar_t *tmpbuf=(wchar_t*)malloc((inbuf+1)*SIZEOF_WCHAR_T);
- memcpy(tmpbuf,psz,(inbuf+1)*SIZEOF_WCHAR_T);
- WC_BSWAP(tmpbuf, inbuf)
- psz=tmpbuf;
-#endif
if (buf)
{
// have destination buffer, convert there
res += 16 - outbuf;
} while ((cres==(size_t)-1) && (errno==E2BIG));
}
-#ifdef WC_NEED_BSWAP
- free(tmpbuf);
-#endif
+
+ if (g_wcNeedsSwap)
+ {
+ free(tmpbuf);
+ }
+
if (ICONV_FAILED(cres, inbuf))
+ {
+ //VS: it is ok if iconv fails, hence trace only
+ wxLogTrace(wxT("strconv"), wxT("iconv failed: %s"), wxSysErrorMsg(wxSysErrorCode()));
return (size_t)-1;
+ }
return res;
}
bool usable()
{ return (m2w != (iconv_t)-1) && (w2m != (iconv_t)-1); }
-public:
+protected:
iconv_t m2w, w2m;
};
#endif
size_t WC2MB(char *buf, const wchar_t *psz, size_t n)
{
-#if defined(__BORLANDC__) && (__BORLANDC__ > 0x530)
+#if ( defined(__BORLANDC__) && (__BORLANDC__ > 0x530) ) \
+ || ( defined(__MWERKS__) && defined(__WXMSW__) )
size_t inbuf = std::wcslen(psz);
#else
size_t inbuf = ::wcslen(psz);
SetName(name);
}
- m_cset = wxGetCharacterSet(m_name);
+ // wxGetCharacterSet() complains about NULL name
+ m_cset = m_name ? wxGetCharacterSet(m_name) : NULL;
m_deferred = FALSE;
}
}
return m_cset->WC2MB(buf, psz, n);
// latin-1 (direct)
-#if defined(__BORLANDC__) && (__BORLANDC__ > 0x530)
+#if ( defined(__BORLANDC__) && (__BORLANDC__ > 0x530) ) \
+ || ( defined(__MWERKS__) && defined(__WXMSW__) )
size_t len=std::wcslen(psz);
#else
size_t len=::wcslen(psz);