=?ISO-8859-1?Q?LastError ( was DBCS)?=

[juergen . schmied]

>   len = MultiByteToWideChar(CP_ACP,0,astr,-1,NULL,0);
I noticed every call with the len = 0 to get the real length of the string sets the 
LastError to ERROR_INVALID_PARAMETER. We should fix that.
Don't know exactly how but I think when internally used MultiByteToWideChar should not 
influence the value of the LastError.

I think we have the same problem with some more functions.

Juergen