sysErrorString are using "A" function.
If OS environment is multibyte, that result include multibyte string.

On Thu, 14 Jul 2005 08:24:36 +0000 (UTC), k2 <k2_member@pathlink.com> wrote:
> sysErrorString are using "A" function.
> If OS environment is multibyte, that result include multibyte string.

This should do the trick, I suspect.

//This code should be considered public domain
wchar[] sysErrorStringW(uint errcode)
{
    wchar[] result;
    wchar* buffer;
    DWORD r;

    r = FormatMessageW(
	    FORMAT_MESSAGE_ALLOCATE_BUFFER |
	    FORMAT_MESSAGE_FROM_SYSTEM |
	    FORMAT_MESSAGE_IGNORE_INSERTS,
	    null,
	    errcode,
	    MAKELANGID(LANG_NEUTRAL, SUBLANG_DEFAULT), // Default language
	    cast(LPWSTR)&buffer,
	    0,
	    null);

    /* Remove \r\n from error string */
    if (r >= 2)
	r -= 2;
    result = buffer[0..r].dup;
    LocalFree(cast(HLOCAL)buffer);
    return result;
}

Regan

The following sample codes get an error, "Error: 4invalid UTF-8 sequence", on my PC:

import std.file;
import std.utf;
import std.c.windows.windows;

extern(Windows) export
int MessageBoxW(HWND hWnd, LPCWSTR lpText, LPCWSTR lpCaption, UINT uType);

void main()
{
try{
std.file.isdir("notexist");
}catch(Object o){
MessageBoxW(null, toUTF16z(o.toString()), null, MB_OK);
}
}

isdir, etc are using sysErrorString.
So, I think this is better:

char[] sysErrorString(uint errcode)
{
char[] result;
wchar* buffer;
DWORD r;

r = FormatMessageW(
FORMAT_MESSAGE_ALLOCATE_BUFFER |
FORMAT_MESSAGE_FROM_SYSTEM |
FORMAT_MESSAGE_IGNORE_INSERTS,
null,
errcode,
MAKELANGID(LANG_NEUTRAL, SUBLANG_DEFAULT), // Default language
cast(LPWSTR)&buffer,
0,
null);

/* Remove \r\n from error string */
if (r >= 2)
r -= 2;
result = std.utf.toUTF8(buffer[0..r]);
LocalFree(cast(HLOCAL)buffer);
return result;
}

On Thu, 14 Jul 2005 09:54:46 +0000 (UTC), k2 <k2_member@pathlink.com> wrote:
> The following sample codes get an error, "Error: 4invalid UTF-8 sequence", on my
> PC:
>
> import std.file;
> import std.utf;
> import std.c.windows.windows;
>
> extern(Windows) export
> int MessageBoxW(HWND hWnd, LPCWSTR lpText, LPCWSTR lpCaption, UINT uType);
>
> void main()
> {
> try{
> std.file.isdir("notexist");
> }catch(Object o){
> MessageBoxW(null, toUTF16z(o.toString()), null, MB_OK);
> }
> }
>
> isdir, etc are using sysErrorString.
> So, I think this is better:
>
> char[] sysErrorString(uint errcode)
> {
> char[] result;
> wchar* buffer;
> DWORD r;
>
> r = FormatMessageW(
> FORMAT_MESSAGE_ALLOCATE_BUFFER |
> FORMAT_MESSAGE_FROM_SYSTEM |
> FORMAT_MESSAGE_IGNORE_INSERTS,
> null,
> errcode,
> MAKELANGID(LANG_NEUTRAL, SUBLANG_DEFAULT), // Default language
> cast(LPWSTR)&buffer,
> 0,
> null);
>
> /* Remove \r\n from error string */
> if (r >= 2)
> r -= 2;
> result = std.utf.toUTF8(buffer[0..r]);
> LocalFree(cast(HLOCAL)buffer);
> return result;
> }

Ahh, of course, FormatMessageA is giving a multibyte result which isn't UTF8. FormatMessageW gives a unicode result AKA UTF-16.

Does it bother anyone else that we do UTF conversions all the time? Here we go from UTF-16 to UTF-8. I noticed in particular doFormat uses a delegate taking a dchar so each and every character formatted gets converted to/from dchar. It just seems inefficient in some way.

Regan