There is a small bug in detecting UTF8 encoding of xml source. Because of this bug utf8 xmls with unicode characters are loaded improperly with IrrXmlReader
---
in function
bool readFile(IFileReadCallBack* callback)
we have
const unsigned char UTF8[] = {0xEF, 0xBB, 0xBF};
and later we comparing BOM with source header
if (size >= 3 && data8[0] == UTF8[0] && data8[1] == UTF8[1] && data8[2] == UTF8[2]) // Bug!
On my MSVC Express this "if" is never triggered, even if source hase this BOM in proper place.
the reason is type casting (works weird here):
- UTF8[] is "unsigned char"
- data8[] is "char" (char* data8 = new char[size])
=> (UTF8[2] = 0xBF) and (data8[2] = -63)!
I would suggest to add explicit type conversion on both sides to avoid any misinterpretaions. Obviously there is no overhead, just compile-time hint
if (size >= 3 && char(data8[0]) == char(UTF8[0]) && char(data8[1]) == char(UTF8[1]) && char(data8[2]) == char(UTF8[2]))
that is all
